Ecology, 93(7), 2012, pp. 1683–1694� 2012 by the Ecological Society of America

Above- and belowground responses of arctic tundra ecosystemsto altered soil nutrients and mammalian herbivory

LAURA GOUGH,1,5 JOHN C. MOORE,2,3 GAUIS R. SHAVER,4 RODNEY T. SIMPSON,3 AND DAVID R. JOHNSON1,6

1Department of Biology, University of Texas, Arlington, Texas 76019 USA2Department of Ecosystem Science and Sustainability, Colorado State University, Fort Collins, Colorado 80523 USA

3Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado 80523 USA4Ecosystems Center, Marine Biological Laboratory, Woods Hole, Massachusetts 02543 USA

Abstract. Theory and observation indicate that changes in the rate of primary productioncan alter the balance between the bottom-up influences of plants and resources and the top-down regulation of herbivores and predators on ecosystem structure and function. TheExploitation Ecosystem Hypothesis (EEH) posited that as aboveground net primaryproductivity (ANPP) increases, the additional biomass should support higher trophic levels.We developed an extension of EEH to include the impacts of increases in ANPP onbelowground consumers in a similar manner as aboveground, but indirectly through changesin the allocation of photosynthate to roots. We tested our predictions for plants abovegroundand for phytophagous nematodes and their predators belowground in two common arctictundra plant communities subjected to 11 years of increased soil nutrient availability and/orexclusion of mammalian herbivores. The less productive dry heath (DH) community met thepredictions of EEH aboveground, with the greatest ANPP and plant biomass in the fertilizedplots protected from herbivory. A palatable grass increased in fertilized plots while dwarfevergreen shrubs and lichens declined. Belowground, phytophagous nematodes alsoresponded as predicted, achieving greater biomass in the higher ANPP plots, whereaspredator biomass tended to be lower in those same plots (although not significantly). In thehigher productivity moist acidic tussock (MAT) community, aboveground responses werequite different. Herbivores stimulated ANPP and biomass in both ambient and enriched soilnutrient plots; maximum ANPP occurred in fertilized plots exposed to herbivory. Fertilizedplots became dominated by dwarf birch (a deciduous shrub) and cloudberry (a perennial forb);under ambient conditions these two species coexist with sedges, evergreen dwarf shrubs, andSphagnum mosses. Phytophagous nematodes did not respond significantly to changes inANPP, although predator biomass was greatest in control plots. The contrasting results ofthese two arctic tundra plant communities suggest that the predictions of EEH may hold forvery low ANPP communities, but that other factors, including competition and shifts invegetation composition toward less palatable species, may confound predicted responses tochanges in productivity in higher ANPP communities such as the MAT studied here.

Key words: Arctic LTER, Toolik Lake, Alaska, USA; arctic tundra ecosystem; biomass; caribou;EEH, exploitation ecosystem hypothesis; food web; herbivory; nematode; nutrients; predators; productivity;vole.

INTRODUCTION

Oksanen et al. (1981) developed the Exploitation

Ecosystem Hypothesis (EEH) to describe how food

chains develop along aboveground net primary produc-

tivity (ANPP) gradients based on relatively low-produc-

tivity subarctic ecosystems supporting vertebrate

herbivores and predators. As ANPP increases from a

minimal level, plant biomass also increases until a

threshold is reached where herbivore populations can be

sustained. At that level of ANPP, plant biomass

stabilizes while ANPP continues to increase; the

additional productivity is consumed by herbivores so

that consumer biomass increases but plant biomass does

not. Similar patterns are predicted for further increases

in ANPP that will add additional trophic levels to the

system.

These patterns predicted by EEH have been validated

in several instances in terrestrial ecosystems (reviewed in

Wardle 2002), although they have not been explicitly

tested along above- or belowground productivity

gradients very often, perhaps because ecologists felt

that EEH was too simplistic. For example, the EEH

system was described as a linear food chain or series of

coupled linear food chains, wherein each trophic level

was treated as a single population and selective

Manuscript received 9 September 2011; revised 20 December2011; accepted 1 February 2012. Corresponding Editor:H. A. L. Henry.

5 E-mail: gough@uta.edu6 Present address: Department of Biological Sciences,

University of Texas, El Paso, Texas 79968 USA.

1683

herbivory was ignored. Additionally, trophic cascades

have often been considered to be weak in terrestrial

ecosystems (e.g., Strong 1992; but see Schmitz et al.

2004). Fraser and Grime (1999) argued that some tests

of EEH were overcriticized because of individual biases

against terrestrial trophic cascades, whereas their

terrestrial mesocosm results point to the existence of

strong trophic cascades that follow the predictions of

EEH. Certainly there have been many tests of the

interactive effects of adding nutrients (increasing pro-

ductivity) and herbivory (e.g., Gruner et al. 2008) in

terrestrial ecosystems including tundra (e.g., Grellmann

2002), but these studies usually have not been examined

in the context of EEH.

Recently, Aunapuu et al. (2008), conducting a series

of studies in a subarctic Scandinavian heath in which the

number of trophic levels or soil nutrients were experi-

mentally manipulated, found support for EEH. When

nutrients were added, plant biomass remained similar to

control levels, whereas rodent herbivore activity and

biomass significantly increased. Vegetation that had

been transplanted to islands with both herbivores and

secondary consumers maintained greater biomass than

vegetation transplanted to islands with only herbivores.

These results confirm that subarctic tundra can behave

according to the assumptions of EEH.

Here we test the predictions of EEH in two common

low-productivity Alaskan tundra communities, dry

heath (DH) and moist acidic tussock (MAT), subject

to manipulations of soil nutrients and mammalian

herbivory for 11 years. MAT and DH tundra differ in

ANPP, biomass, plant species composition, response to

added nutrients, and herbivore activity (e.g., Shaver et

al. 2001, Gough et al. 2002, 2007). We developed an

extension of EEH to simultaneously study aboveground

and belowground responses to changes in plant produc-

tivity and herbivory using the framework presented by

Schmitz (2008). Such studies that simultaneously exam-

ine soil fauna, vegetation, and mammalian herbivores

aboveground are quite rare (reviewed in Bardgett and

Wardle 2010), yet are crucial to better understand how

arctic ecosystems function (Post et al. 2009, Wookey et

al. 2009).

We predicted that both aboveground and below-

ground responses to manipulations would result from a

combination of direct and indirect effects of production

and consumption, with the magnitude of the responses

being dependent on the allocation of production within

the plants to shoots and roots. Because productivity in

our study system is low, we expect that two trophic

levels dominate aboveground interactions (predators are

relatively rare). We also assume that because our

experimental plots are relatively small, altering produc-

tivity in situ will not affect population structure of

herbivores at the landscape scale, but will affect activity

of herbivores in the plots themselves. Our aboveground

predictions (denoted ‘‘A’’) follow directly from EEH

(Fig. 1A):

H1A: When nitrogen and phosphorus (þNP) are

added to the soil, ANPP increases, but biomass does not

because herbivores consume the additional productivity.

H2A: When mammalian herbivores are removed

(�H), aboveground biomass increases because some of

the productivity will now be able to accumulate as

standing crop, but ANPP will not change.

H3A: When nutrients are added and herbivores are

removed (þNP�H), ANPP and biomass exceed the

combined levels of the other two treatments alone (a

superadditive interaction, sensu Gruner et al. 2008). This

assumes that herbivory and nutrients interact such that

mammals are attracted to fertilized plots; protecting

fertilized plants from herbivory allows the plant

community to increase both ANPP and biomass.

Our belowground predictions (denoted ‘‘B’’; Fig. 1B)

incorporate both belowground herbivores and their

FIG. 1. Hypothesized responses of (A) aboveground plantbiomass and (B) belowground consumer biomass (herbivoresand their predators) to changes in productivity (ANPP) causedby experimental treatments in the Arctic LTER, Toolik Lake,Alaska, USA. HA and HB, respectively, are hypotheses foraboveground and belowground responses. Treatment abbrevi-ations are: CT, control; þNP, added soil nitrogen andphosphorus; �H, mammalian herbivores excluded. The H2B(arrow from CT to�H, where B indicates belowground) is notshown as no change is predicted from control levels. SeeMethods: Study site and experimental treatments for details.

LAURA GOUGH ET AL.1684 Ecology, Vol. 93, No. 7

predators and are based on how our experimental

manipulations are predicted to alter ANPP in accor-dance with EEH for a system with up to four trophic

levels, and on expected changes in the allocation ofproductivity between shoots and roots. We assume from

previous studies in these communities (e.g., Gough et al.2002, van Wijk et al. 2003, Hobbie et al. 2005) thatfertilization will increase ANPP and root production,

but that removal of mammalian herbivores alone (�H)will not have an effect on root production. The

hypotheses for belowground response are:H1B: In fertilized plots, belowground herbivore

biomass (phytophagous nematodes) and predator bio-mass increase.

H2B: When mammals are removed, no change in thebiomass of belowground herbivores or predators occurs.

H3B: When nutrients are added and herbivoresremoved, belowground herbivore biomass and predator

biomass increase more than under H1B.

METHODS

Study site and experimental treatments

This research was conducted at the Arctic Long-

Term Ecological Research (LTER) site at Toolik Lake,Alaska (68.28 N, 149.68 W, 760 m a.s.l.). In 1996 a

factorial design was implemented in both MAT andDH to determine how the presence or absence ofmammals affected vegetation and soil responses to

added nutrients (10 g�m�2�yr�1 as NH4NO3 and 5g�m�2�yr�1 as P2O5 applied annually in granular form in

early June following snowmelt). Treatment plots of 53

20 m were replicated within four blocks at MAT and

three blocks at DH. For this experiment, one plotwithin each block was randomly assigned N and P

addition (þNP) and another had no nutrients added asa control (CT).

Half of each 5 3 20 m plot in each block was leftunfenced so that mammals could easily access the

vegetation, while the remaining 5 3 10 m was enclosedin a large-mesh fence designed to exclude caribou (15.2

315.2 cm openings, ;1.2 m in height). Half of the area(5 3 5 m) within the 5 3 10 m fence was additionally

enclosed by a small-mesh fence (1.33 1.3 cm openings,;0.8 m in height). The small-mesh fence was buried in

the soil at least 10 cm at construction to preventanimals from burrowing into the plots and wasdesigned to additionally exclude small mammals

including ground squirrels, voles, and lemmings. Forthis study, we did not consider the effects of small and

large mammals separately; thus all data were collectedfrom unfenced areas and areas with both small- and

large-mammal exclusion (designated as�H for lackingherbivory). Each block thus contained plots with four

possible combinations of fence and fertilization: CT,þNP, �H, and þNP�H. Although there are herbivo-

rous insects in these communities, their effect on leafbiomass appeared to be minimal (L. Gough and D. R.

Johnson, personal observations). These plots may

retain snow on the immediate northern edge of fences

for 2–3 days early in the season relative to unfenced

areas (L. Gough and D. R. Johnson, personal

observations). To avoid this snow accumulation, all

sampling was conducted at least 0.5 m from the edge of

the fence.

Aboveground communities

Mammalian herbivores.—Five species of microtine

rodents have been recorded on the North Slope of

Alaska (Batzli et al. 1980): three species of voles

(Microtus oeconomus, M. miurus, and Clethrionomys

rutilis) and two lemming species (Dicrostonyx rubricatus

and Lemmus trimucronatus). The tundra vole (M.

oeconomus) and singing vole (M. miurus) are common

in communities near the Arctic LTER (Batzli and

Lesieutre 1995). Specifically, tundra voles are commonly

seen at MAT (L. Gough and D. R. Johnson, personal

observations), as is evidence of their presence (hay piles,

trails, and fecal deposits) during cyclical outbreak years.

Singing voles and collared lemmings (D. rubricatus) have

been found in rocky areas very near DH (Batzli and

Henttonen 1990; L. Gough and D. R. Johnson, personal

observations). Batzli and Henttonen (1990) suggested

that density of these rodents is limited by food

availability but may also be restricted top-down by

predators similar to lemmings in coastal tundra (Batzli

et al. 1980).

Although they are transient, caribou are commonly

sighted near the Arctic LTER (L. Gough and D. R.

Johnson, personal observations). Toolik Lake lies within

the range of the Central Arctic Herd (Lenhart 2002), but

the herd’s primary calving grounds are far to the north.

Thus caribou are not thought to be common foragers of

plants in MAT. In DH however, caribou feces are

frequently seen following snowmelt. Thus, DH may be

an important winter habitat for caribou because snow

cover is often less there than in surrounding areas

(Cheng et al. 1998; see also Gough et al. 2008).

Mammalian activity was estimated in all plots in an

area of 8 m2 when vegetation was sampled nondestruc-

tively for relative abundance in late July 2006. Observers

noted the percentage of ground surface on which signs

of activity were seen, including vole trails, hay piles, and

fecal piles, evidence of litter that had been chewed by

mammals, and caribou fecal pellets.

Plant biomass

To estimate ANPP and leaf and shoot biomass, we

harvested four 20 3 20 cm quadrats located randomly

along a 5-m transect in each treatment plot in each block

in DH and in three blocks in MAT in late July 2006.

Each quadrat was cut 10–20 cm deep to sample all

rhizomes, and we considered aboveground plant bio-

mass ‘‘within’’ the quadrat if plants were associated with

a meristem inside the quadrat. We sorted all vascular

plants to species in a field laboratory and lumped mosses

and lichens. Vascular species were then sorted into tissue

July 2012 1685ABOVE- AND BELOWGROUND TEST OF EEH

type, including new leaves, new stems, old leaves, old

stems, and rhizomes (as appropriate for particular

growth forms). All samples were dried at 608C to a

constant mass and were weighed to the nearest

milligram. ANPP was determined separately by growth

form. Graminoid and forb ANPP was calculated as the

sum of all leaves and aboveground stems that had been

produced that year. Deciduous and evergreen shrub

ANPP was calculated by summing new leaf mass, new

stem mass, and an estimate of stem secondary growth

calculated as a percentage of old stem mass following

relationships for control andþNP treatments established

in Bret-Harte et al. (2002). Moss and lichen productivity

was not determined.

We took a separate set of samples to estimate root

biomass. At DH, a 5 cm diameter soil corer was used to

extract a soil core adjacent to each location where a

quadrat was taken as described previously; maximum

core depth was 10 cm, as no roots were seen below that

depth. At MAT, a 53 10 cm piece of tundra was cut out

adjacent to each aboveground sampling quadrat to the

depth of permafrost. In the laboratory each sample was

separated into organic and mineral horizons and then

was sliced into 5-cm increments. Roots were extracted

from each section, carefully washed, and separated to

genus or species when possible before being dried for at

least three days at 608C and weighed. Because sampling

roots from all plots was not logistically possible, we only

sampled one block of �H and þNP�H treatments in

each community.

All plant tissues were shipped back to the University

of Texas at Arlington where they were dried again,

lumped by block, and then analyzed for C and N

content using a Perkin Elmer Series 2200 CHN Analyzer

(Perkin Elmer, Waltham, Massachusetts, USA).

Belowground communities

The organic horizons and upper 5 cm of the mineral

horizon were sampled in MAT and DH in late July

2006. In control plots, the organic horizon at MAT

often reaches 15 cm in depth, whereas at DH it averages

2–3 cm depth (J. C. Moore, unpublished data). In MAT

we stratified our sampling to include tussock areas,

taken immediately adjacent to a tussock of the sedge

Eriophorum vaginatum, and inter-tussock areas, mossy

depressions between the tussocks, because of the distinct

microtopographic differences between the two. Because

E. vaginatum covers approximately one-third of the

ground surface in MAT control plots (Hobbie et al.

2005), we weighted the paired tussock samples at 1/3

and the inter-tussock samples at 2/3 to determine

abundance at the square-meter scale. In DH samples

were taken at random because the plots were carpeted

with lichens, with no discernible difference in micro-

topography.

In both MAT and DH, two 10310 cm columns of the

organic horizon through the upper 5 cm of the mineral

soil layer were taken in each sampling location. The

sample pairs from each plot were sliced in half, with one

half kept intact and used to extract microarthropods and

the other half passed through an 8-mm sieve for

additional sampling (not presented here). All biomass

estimates are reported on a ‘‘per gram of dry soil’’ basis

before being converted to mg C/m2. Dry mass values for

soils used to extract microarthropods were estimated by

the difference between the pre- and post-extraction mass

of the soils from which the microarthropods were

extracted, as this process involved heating the soils from

above until they were dry.

Invertebrate biomass.—Biomass (mg C/m2) was esti-

mated from the counts and morphometric conversions

for each functional group assuming 50% C content and

an estimate of the bulk density of the soil (Hunt et al.

1987, Doles 2000). The functional groups reported here

include phytophagous nematodes and nematode preda-

tors extracted from soils using techniques that will be

described. To estimate biomass of the predators of the

phytophagous nematodes, the biomass of predatory

nematodes, predatory tardigrades, and predatory mites

(nematophagous Mesostigmata, Prostigmata, and As-

tigmata) were summed within each soil sample.

Nematodes were isolated from 20 g of sieved soil

using Baermann funnels (Baermann 1917). After three

days on the apparatus, nematodes that were collected

were preserved in formalin and sorted into functional

groups. The phytophagous nematodes (plant-feeding

and root-associate) and predatory nematodes are

reported here. The phytophagous nematodes, members

of the family Tylenchidae, were subdivided into free-

living plant feeding nematodes and root-associate

nematodes using the criteria of Yeates et al. (1993)

based on the morphology of their mouthparts and

feeding habits. The mouthparts of free-living plant

feeding nematodes possess robust stylets for feeding on

plant roots, root hairs, and rhizomes, whereas root-

associate nematodes possess fine stylets and feed on root

epidermal cells, root hairs, lichens, fungi, and algae.

Tardigrades were isolated from 5 g of sieved soil

placed in a 100 mm diameter petri dish and immersed in

deionized water. The submerged soils sat for 5 minutes

at room temperature after which coarse organic debris

was rinsed to release the tardigrades into the dish; the

organic debris was then removed. The dishes were

placed under a dissecting microscope and live tardi-

grades were counted.

Microarthropods were heat-extracted into 90% etha-

nol, 1% glycerine from soils using Tullgren funnels

(Moore et al. 2000). The intact sample pairs from each

plot were wrapped together in cotton cheesecloth,

weighed, placed over a funnel, and heated with a 9-W

incandescent lightbulb for 5 days until dry. The dry soil

and cheesecloth were reweighed to determine the dry

mass of soil per sample and the soil moisture content.

The collected arthropods were sorted into taxonomic

and functional groups (sensu Moore et al. 1988).

LAURA GOUGH ET AL.1686 Ecology, Vol. 93, No. 7

Statistical analyses

Analysis of variance was used to assess plant and soil

organism responses to the fertilization and grazing

treatment. All statistical tests were run using SAS

(Version 9.2, SAS Institute, Cary, North Carolina,

USA). For all variables, results were analyzed separately

for DH and MAT communities. Within each commu-

nity, a two-way randomized-block design ANOVA was

conducted with fertilization and herbivory as the main

effects, quadrat nested in block, and block included. For

the microarthropods and plant tissue analysis, the

design did not include the nested term because the

samples were bulked for extraction or analysis. Data

were natural log-transformed as needed to meet

parametric model assumptions. Tukey’s post hoc

comparisons were used to determine significant pairwise

differences. All data are reported as mean 6 SE.

RESULTS

Mammal activity

Adding nutrients resulted in an increase in mamma-

lian activity in both communities. At DH, caribou feces

were evident in open plots (3% 6 1% in CT and 1% 6

0% in þNP). Caribou litter (pieces of the grass

Hierochloe alpina that had been removed from the plant

via trampling) was present in 11% 6 6% of the area of

þNP plots. Evidence of vole hay piles was also found in

the þNP plots.

At MAT, recent vole activity was seen in the open

plots, with 2% 6 1% of the area of CT plots and 4% 6

2% of þNP plots affected as evidenced by vole fecal

pellets, runways, and litter. Eriophorum tussocks showed

direct signs of vole damage including hay piles and

tunnels, with more extreme effects, e.g., killing of

sections of or entire tussocks by winter consumption

of rhizomes and leaf bases, observed in the þNP plots.

Plant productivity and tissue nutrient content

In DH, the pattern of ANPP across treatments

followed predictions (see Plate 1), but there was no

significant effect of any treatment on total vascular

ANPP (Fig. 2A). Fertilization resulted in significantly

lower evergreen ANPP (F1,31 ¼ 46.91, P , 0.0001) and

greater graminoid ANPP, primarily Hierochloe (F1,33 ¼37.87, P , 0.0001; Fig. 2A). Evergreen shrub ANPP was

also marginally higher in plots protected from herbi-

vores (F1,31¼ 3.52, P¼ 0.07), suggesting a benefit to this

group from the exclosures. Fertilization decreased the

C:N ratio of total vascular ANPP (F1,8 ¼ 253.11, P ,

0.0001), while excluding herbivores had the opposite

effect and increased C:N (F1,8¼ 9.32, P¼ 0.02; Table 1).

In contrast, vascular ANPP at MAT was significantly

greater in plots subject to fertilization (F1,31¼56.56, P ,

0.0001) and those exposed to herbivores (F1,31¼ 4.95, P

¼ 0.03), with no interaction between these main effects

(Fig. 2B). A herbivore effect was also not detected for

ANPP of any of the growth forms when analyzed

separately (Fig. 2B). As at DH, evergreen shrubs at

MAT had lower ANPP in fertilized plots (F1,32¼ 22.48,

P , 0.0001), whereas deciduous shrubs and forbs had

greater ANPP in those plots (F1,32 ¼ 17.09, P ¼ 0.0002

and F1,33¼13.78, P¼0.0008, respectively). Although the

graminoid ANPP means were different (Fig. 2B),

substantial within-treatment variation precluded the

detection of any significant effects. The C:N of total

vascular ANPP was also significantly reduced by

fertilization (F1,8 ¼ 28.28, P ¼ 0.0007), but showed no

effect of herbivore exclosure (Table 1).

Plant biomass

At DH, total biomass was significantly greater inside

herbivore exclosures than outside (F1,29¼ 4.83, P¼ 0.04;

Fig. 3A). This pattern was driven by positive responses

to herbivore exclosure by mosses and deciduous shrubs

(marginally significant, P , 0.10). Woody stem biomass

(summed across all woody species; data not shown) was

also marginally greater inside exclosures (F1,31¼ 3.66; P

FIG. 2. Aboveground net primary productivity (ANPP;mean 6 SE) after 11 years of experimental manipulation in (A)dry heath tundra and (B) moist acidic tussock tundra.Treatment abbreviations are as in Fig. 1. See Results forstatistical results by growth form.

July 2012 1687ABOVE- AND BELOWGROUND TEST OF EEH

¼ 0.06), probably reflecting subtle responses by shrubs

to long-term removal of herbivore pressure.

Fertilization induced significantly less total live

biomass at DH (F1,29 ¼ 6.95, P ¼ 0.01), probably due

to its significant negative effects on evergreen and lichen

biomass (F1,31¼ 29.81, P , 0.0001 and F1,33¼ 70.33, P

, 0.0001, respectively; Fig. 3A). Graminoid biomass

was significantly greater in fertilized plots (F1,33¼ 37.61,

P , 0.0001), but this was not enough to offset the losses

of other groups. Woody stem biomass was greater in

control than fertilized plots (F1,31 ¼ 5.00, P ¼ 0.03),

suggesting that fertilization caused a loss of wood

biomass at DH.

Root biomass at DH was greater in fertilized plots

(F1,17¼7.43, P¼0.01; Fig. 4A) and extended deeper into

mineral soil (data not shown). This partly resulted from

greater biomass of grass (Hierochloe) roots inþNP plots

(Fig. 4A). There were no significant effects of herbivory

on root biomass. To a lesser extent than for ANPP,

fertilization reduced overall root C:N (F1,4¼ 11.15, P¼0.03), but mammal exclosures had no effect (Table 1).

At MAT there were also no significant interactions

between herbivory and nutrient addition for any of the

biomass variables examined. In contrast to DH,

aboveground biomass (both including mosses and

lichens and when restricted to vascular species) was

significantly lower inside herbivore exclosures (F1,32 ¼4.50, P ¼ 0.04 and F1,32 ¼ 5.27, P ¼ 0.03, respectively;

Fig. 3B). Fertilization as a main effect caused more

dramatic responses. Aboveground vascular biomass and

vascular total biomass were greater in fertilized plots

relative to controls, driven by greater deciduous shrub

and forb (Rubus chamaemorus) biomass (all P , 0.05).

In particular, aboveground biomass of the dwarf

deciduous shrub Betula nana was significantly greater

in fertilized plots (F1,33 ¼ 14.82, P ¼ 0.0005). Other

groups were negatively affected; evergreen shrubs,

mosses, and lichens had less biomass in the fertilized

plots (all P , 0.05). In contrast with DH, woody stem

biomass was greater in fertilized plots (F1,31¼ 30.02, P¼0.0001).

The response of root biomass to fertilization and

herbivore exclusion was variable in MAT (Fig. 4B).

Total root biomass was not significantly affected by any

TABLE 1. Carbon to nitrogen (C:N) ratios for new shoot tissues (new leaves and new stems) and root tissues for the entire plantcommunity sampled in each quadrat or soil core after 11 years of experimental treatment in the Arctic LTER, Toolik Lake,Alaska, USA.

Tissue Site

C:N ratio, by treatment (mean 6 SE)

CT þNP �H þNP�H

New shoot DH 31a 6 1 12b 6 1 36a 6 1 14b 6 2MAT 24a 6 2 17b 6 1 24a 6 2 17b 6 1

Root DH 36 6 3 24 6 3 36 20MAT 34 6 5 27 6 4 38 25

Notes: Sites and sample sizes are dry heath (DH; n¼ 3 blocks) and moist acidic tussock (MAT; n¼ 4 blocks), except where noerrors are presented (n¼1 block); seeMethods for details. Treatment abbreviations are : CT, control;þNP, added soil nitrogen andphosphorus; �H, mammalian herbivores excluded. Within each row, different superscript letters indicate that the means aresignificantly different at P , 0.05.

FIG. 3. Above- and belowground plant biomass (mean 6SE for vascular plants aboveground, total abovegroundbiomass, and vascular plants belowground) after 11 years ofexperimental manipulation in (A) dry heath tundra and (B)moist acidic tussock tundra. The x-axis represents the groundsurface; on the y-axis, values above and below the x-axis,respectively, indicate aboveground and belowground rhizomebiomass; root biomass is presented in Fig. 4. Treatmentabbreviations are as in Fig. 1. See Results for statistical resultsby growth form.

LAURA GOUGH ET AL.1688 Ecology, Vol. 93, No. 7

treatment, although the sedge Eriophorum had margin-

ally significantly greater biomass in fertilized plots (F1,17

¼ 3.86, P ¼ 0.06). Total root C:N appeared lower in

fertilized plots, although not significantly (Table 1).

Soil herbivores and predators

In DH, soil herbivores responded as predicted to

changes in ANPP. Phytophagous nematodes (plant-

feeding and root-associate forms) had greater biomass in

all treatments relative to control values (main effect of

grazing and fertilization were marginally significant; P

, 0.10); the greatest herbivore biomass occurred in the

þNP�H treatment, where ANPP was also greatest

(Table 2). Within the phytophagous nematode grouping,

the root-associate nematode biomass was greater in

fertilized plots (F1,13 ¼ 4.03, P ¼ 0.06). The addition of

soil nutrients appeared to suppress predator biomass

(predatory nematodes, predatory mites, and tardi-

grades), although not significantly (Table 2). However

within the predator functional group, the tardigrades

exhibited a grazing effect (F1,14 ¼ 4.74, P ¼ 0.05) with

lower biomass in plots from which grazers were excluded

(Table 2).

At MAT, phytophagous nematode biomass was

similar to that in control plots at DH but showed no

response to fertilization or grazing and subsequent

changes in ANPP (Table 2). A significant interaction

between fertilization and grazing affected belowground

predator abundance (F1,12 ¼ 7.52, P ¼ 0.02), as did the

main effects of fertilization (F1,12¼ 11.16, P¼ 0.02) and

grazing (F1,12 ¼ 7.65, P ¼ 0.02), with lower biomass in

the experimental manipulations relative to control plots

(Table 2). Within the predators, fertilization reduced the

biomass of tardigrades (F1,12 ¼ 4.34, P ¼ 0.06). The

biomass of nematophagous mites was lower in the

grazing exclosures (F1,9 ¼ 3.48, P ¼ 0.10).

DISCUSSION

In our study, as in similar simultaneous manipulations

of herbivore pressure and nutrient availability in other

terrestrial ecosystems (see meta-analysis by Gruner et al.

2008), fertilization and herbivore exclosure generally did

not interact statistically to affect ANPP or biomass.

These findings run counter to the EEH predictions that

assume that with increased soil nutrients and ANPP,

effects of herbivory should be more pronounced. That

said, we did see trends suggesting such interactions, but

variability in response within treatments was frequently

too large to detect significant effects.

Dry heath responded as predicted by EEH

When presented graphically, aboveground biomass in

the DH plots appeared to respond as predicted by EEH

(Fig. 5A), demonstrating that productivity increases

caused by greater soil nutrient availability were only

manifested as increases in biomass when mammals were

excluded from the plots (supporting H1A [þNP] and

H3A [þNP�H]). However, vascular biomass did not

significantly increase as a result of herbivore exclosure

alone; thus H2A (�H) was not supported. These results

suggest that herbivores are not very important in

ambient nutrient plots, but are attracted to fertilized

plots in this community (also see Grellmann 2002). Not

only does the abundance of the highly palatable grass,

Hierochloe, increase, but also overall C:N of ANPP

declines and less wood is produced in fertilized plots.

These changes together suggest that with increased soil

nutrients at DH, both quantity and quality of forage

improve for mammals.

Herbivores have been shown to stimulate plant

productivity in other arctic systems by promoting

growth of graminoids (Olofsson et al. 2004, van der

Wal and Brooker 2004, van der Wal 2006) and may

convert heathland to grassland (Olofsson et al. 2001,

2004, Stark et al. 2002). Those European communities

have a significant moss component, so that one

mechanism of stimulation is via trampling by herbi-

vores, which negatively affects mosses, resulting in

increased soil temperatures that benefit graminoids

(Gornall et al. 2009). In the DH tundra studied here,

mosses do not occur in ambient nutrient plots, so other

aspects of animal activity, such as local concentration of

nutrients via feces or urine (van der Wal et al. 2004),

FIG. 4. Root biomass (mean 6 SE) after 11 years ofexperimental manipulation; values below the x-axis simplyindicate belowground measurements. In �H and þNP�Htreatments, roots were only sampled from one block so thereare no error bars. Note the difference in y-axis scales betweensites. See Results for statistical results by growth form.

July 2012 1689ABOVE- AND BELOWGROUND TEST OF EEH

must be involved in the graminoid stimulation that we

detected.

The belowground herbivores in DH also responded as

predicted by EEH but with high variability (Fig. 5B).

This high variability was not entirely unexpected, given

the complex nature of the belowground interactions

(i.e., potential number of trophic levels and feeding

pathways). For example, if we trace the pathways from

roots to phytophagous nematodes to predators using the

functional groups that we have designated here, the

system has as few as three trophic levels. If we

disaggregate the predators into the separate functional

groups (predatory nematodes, tardigrades, and preda-

tory mites; see Table 2) and retrace the paths, there are

between three and six trophic levels due to intraguild

predation, yielding an average and median number of

levels of 4.5. If we focus on the dominant predator

groups (predatory mites and tardigrades) based on our

biomass estimates, and retrace the paths, the mean and

median number of trophic levels is 4 (also see Moore

and de Ruiter 2012, in press).

Under the assumption of four trophic levels, the

caveats previously discussed notwithstanding, phytoph-

agous nematodes appeared to follow trends in ANPP

more closely than patterns of root biomass. Specifically,

the root-associate nematodes appeared to increase in

biomass as a result of improved forage quality

(decreased C:N) and quantity in fertilized plots,

consistent with De Deyn et al. (2004). Predators,

however, had similar or lower biomass in all treatments

compared with control plots (Fig. 5C), with the lowest

biomass in þNP�H where the greatest phytophagous

nematode biomass was recorded. One possible explana-

tion for this result is that the conversion to a grassland

in the DH creates an inhospitable environment for these

predators or shifts feeding patterns by other inverte-

brates, thus depriving this group of a key source of prey.

Another possibility is that the predators faced increased

competition and intraguild predation under these

experimentally induced conditions. Because the bio-

masses of the dominant group (tardigrades and preda-

tory arthropods) changed as a proportion of total

predator biomass (Table 2), the changes in phytopha-

gous nematode densities and predator densities are

indicative of a shift in the balance between bottom-up

and top-down controls on trophic structure consistent

with predictions of EEH.

At MAT mammalian herbivores stimulated ANPP

and biomass

In contrast with DH, MAT productivity and biomass

in both ambient and fertilized plots were stimulated by

the presence of herbivores (Fig. 5A), supporting studies

in other systems of overcompensation (e.g., Hilbert et al.

1981, Hik and Jefferies 1990, Leriche et al. 2001) and

counter to EEH and related hypotheses (Gruner et al.

2008). The MAT responses were driven by different

species, but primarily dwarf birch, B. nana. Betula is

known to contain numerous phenolic compounds in its

leaves, ostensibly making it less desirable to herbivores

(Graglia et al. 2001), but has also responded positively

to reindeer/caribou exclosure in other parts of the Arctic

(e.g., Pajunen et al. 2008, Post and Pedersen 2008,

Olofsson et al. 2009). To our knowledge, our results are

some of the first to suggest a potential stimulation of

Betula by the presence of mammalian herbivores, either

indirectly or directly.

Belowground, biomass of phytophagous nematodes

did not respond to experimental treatments at MAT,

just as root biomass did not (in contrast with Sullivan et

al. 2007), despite the large shifts in aboveground

productivity and root C:N caused by the experimental

treatments (Fig. 5B). Total predators, however, exhib-

ited lower biomass in all the treatment plots at MAT,

similar to the response at DH (Fig. 5C). This might be

related to the increase in Betula and Rubus and the

demise of Eriophorum and Sphagnum mosses. Betula

nana possesses more structurally recalcitrant roots than

Eriophorum and lower quality plant litter (higher C:N)

than co-occurring species (Hobbie 1996), produces

secondary phenolic plant compounds (Graglia et al.

2001), and is highly ectomycorrhizal (Clemmensen et al.

2006). The observed shift in the plant community

essentially transformed the belowground resource base

TABLE 2. Belowground consumer biomass (mean 6 SE) at the two site types.

Belowground consumers

Consumer biomass in DH (mg C/m2)

CT þNP �H þNP�H

Plant-feeding nematodes 0.052 6 0.032 0.151 6 0.126 0.230 6 0.117 0.378 6 0.240Root-associate nematodes 0.035 6 0.017 0.054 6 0.054

Total phytophagous nematodes 0.052 6 0.032 0.185 6 0.109 0.230 6 0.117 0.432 6 0.294

Predatory mites 17.5 6 13.9 10.6 6 9.7 18.0 6 10.0 3.9 6 0.8Nematophagous mites 0.3 6 0.2 0.2 6 0.1 0.1 6 0.1 0.1 6 0.1Predatory nematodesTardigrades 8.0 6 6.1 1.8 6 1.8 0.0 6 0.0 0.7 6 0.7

Total predators 25.8 6 12.5 12.6 6 9.0 18.1 6 10.0 4.7 6 0.3

Notes: Sample size is n¼3 blocks for DH; n¼4 blocks for MAT (see Results: Soil herbivores and predators). Abbreviations are asin Table 1. Within each row, different superscript letters indicate significantly different means at P , 0.05; see Results forstatistically significant main effects. Blank cells indicate that a particular consumer was not found for the specified treatment.

LAURA GOUGH ET AL.1690 Ecology, Vol. 93, No. 7

in ways that may not provide suitable forage for soil

organisms adapted to living within and feeding on

sedges or mosses.

Interacting effects of changes in ANPP, species

composition, tissue quality, and herbivory

One of the most important findings of the work

reported here is how distinctly both nutrients and

mammalian herbivores affected these two communities,

with DH following the predictions of EEH and MAT

responding quite differently. Although we only focused

on two sites for this research, they are representative of

these community types in northern Alaska and thus we

are confident that the differences between the two would

occur at a larger scale. Several reasons seem to be at

play. First, the disparity between aboveground results in

the two communities may stem from the presence of

different herbivores and different plant growth forms in

the two communities. At DH, mammalian herbivores do

not seem to play an important role unless soil nutrients

are added, at which point the newly abundant grasses

provide important resources to caribou and small

mammals. In contrast, at MAT, caribou are transient,

whereas small mammals are important herbivores in

years of high abundance regardless of soil nutrients. Our

results suggest that mammals at MAT may be stimulat-

ing deciduous shrub growth by reducing competition for

nutrients for Betula, because voles selectively target

Eriophorum for food and shelter. A similar competitive

release result has been seen in a removal experiment

conducted in the same community (Bret-Harte et al.

2008).

Second, EEH addresses how an ecosystem responds

along a gradient of productivity. At extremely low

productivity, trophic levels are added with increased

productivity in a stair-stepped fashion. With further

increases in productivity, in the absence of the addition

of a new upper trophic level, biomass is eventually

transferred to upper levels. These changes in structure

are not gradual or linear along the gradient. Rather,

they exhibit a stair-stepped response with the addition of

new trophic levels and sigmoid response in terms of the

accumulation of total biomass, with the more pro-

nounced changes occurring at the low end of the

productivity gradient (Moore et al. 2003). At ;50

g�m�2�yr�1, the DH community is at the low end of the

global ANPP range where significant changes in

ecosystem structure with increased ANPP would be

expected (Moore et al. 1993). At ;150 g�m�2�yr�1, theMAT community is positioned on the response curve

where the observed response to increased ANPP at the

MAT sites would be proportionally less than that

observed in the DH sites; our results confirm these

predictions.

Third, interpreting the results strictly in terms of EEH

is problematic as the hypothesis assumes, if not

implicitly, that the species traits in the community

FIG. 5. Responses of aboveground productivity and (A)aboveground plant biomass, (B) root-feeding nematode bio-mass, and (C) belowground predator biomass to 11 years oftreatment. Solid symbols and solid lines are for DH (dry heath);open symbols and dashed lines are for MAT (moist acidictussock); error bars indicate 6SE.

TABLE 2. Extended.

Consumer biomass in MAT (mg C/m2)

CT þNP �H þNP�H

0.034 6 0.015 0.016 6 0.005 0.046 6 0.036 0.023 6 0.0140.007 6 0.007 0.009 6 0.009

0.034 6 0.015 0.023 6 0.009 0.056 6 0.045 0.023 6 0.014

57.4 6 21.5 3.7 6 1.8 8.8 6 2.0 4.3 6 2.91.4 6 0.9 0.2 6 0.1 0.1 6 0.1 0.2 6 0.20.6 6 0.4 0.2 6 0.2 0.2 6 0.27.6 6 2.6 1.3 6 1.3 4.6 6 2.0 2.9 6 1.5

69.8a 6 21.1 5.7b 6 0.8 13.7b 6 2.3 6.2b 6 2.1

July 2012 1691ABOVE- AND BELOWGROUND TEST OF EEH

remain constant. Both the DH and MAT systems

exhibited shifts in the aboveground communities with

significant declines in species richness caused by

fertilization (Johnson 2008), but did so in ways that

elicited different ecosystem responses. As previously

described, nutrient additions and the removal of

herbivores in DH precipitated a shift in the plant

community from one dominated by lichens and ever-

greens to one with abundant grass, whereas at MAT the

community changed from one of mixed growth forms

(including Sphagnum mosses) to dominance by a dwarf

deciduous shrub and a perennial forb. Another impor-

tant difference between the two sites was that in DH the

increased ANPP accompanied a loss of woody tissues

and a large decrease in C:N, whereas at MAT more

wood was produced, resulting in a proportionally

smaller decrease in C:N. The changes in ANPP at DH

thus follow expected changes along a natural produc-

tivity gradient of an increase in plant productivity and

palatability, whereas at MAT there was less of an

improvement in tissue quality. Evidence suggests that

plant tissue quality can partly control trophic structure

of an ecosystem (Cebrian et al. 2009); thus the different

degree of changes in this factor is likely to partly explain

our overall results.

Finally, we did not see a clear correlation between

changes in belowground herbivores and their predators

and aboveground responses to our treatments. There are

multiple reasons why soil communities may not respond

to increases in ANPP as predicted by EEH (reviewed in

Allison 2006). The suggestion from our study that at

DH phytophagous nematodes increased with fertiliza-

tion while they did not respond in MAT probably also

resulted from changes in the plant community and the

associated rhizosphere. Future studies will continue to

shed light on patterns of belowground consumers in

response to changes in ANPP.

In conclusion, our results demonstrate that the tenets

of EEH can hold in low-productivity terrestrial ecosys-

tems such as the DH studied here, but that in higher-

productivity communities, responses to changes in

ANPP are more complex. In many ways, elements of

the ideas of trophic control and regulation presented by

Hairston et al. (1960), the theory of biogeography (sensu

MacArthur and Wilson 1967), and the strategy of

ecosystem development (Odum 1969) seem appropriate

here. The responses within DH, particularly below-

PLATE 1. Installing an arthropod pitfall trap in early June 2006 before leaf green-up in a fertilized dry heath treatment plot. Theplot is dominated by grasses as a result of long-term nutrient addition; the fence in the background is one of the mammalianherbivore exclosures. Photo credit: R. T. Simpson.

LAURA GOUGH ET AL.1692 Ecology, Vol. 93, No. 7

ground responses, typify those that are bottom-up

driven and resemble the transitions in communities that

occur during the initial developmental phases where

available space allows colonization to outstrip extinc-

tion, and the processes associated with the linkages

among species are quite fluid. The responses in MAT

resemble a community at later stages of development

where species turnover occurs and competition and

predation play a more important role. These findings

suggest that as climate change proceeds in areas such as

arctic tundra, community-level processes that include

higher trophic level responses must be incorporated into

observations, experiments, and models to best predict

community and ecosystem changes.

ACKNOWLEDGMENTS

We gratefully thank Jim Laundre for long-term maintenanceof these experiments and the many field and lab assistantsinvolved in this project, including: Amie Treuer, CarolMoulton, Charla Jordan, Jef Knight, Brian Moon, Karl Wyant,and Greg Selby. Funding for this research was provided by theNational Science Foundation (OPP-0425827 to L. Gough,OPP-0425606 to J. C. Moore, and multiple grants to the MarineBiological Laboratory in support of the Arctic LTER, includingDEB-0423385). Logistic support was provided by Toolik FieldStation, University of Alaska, Fairbanks, and VECO PolarServices. Use of the CHN analyzer for plant tissues wasprovided by Tom Chrzanowski.

LITERATURE CITED

Allison, S. D. 2006. Brown ground: a soil carbon analogue forthe green world hypothesis? American Naturalist 167:619–627.

Aunapuu, M., J. Dahlgren, T. Oksanen, D. Grellmann, L.Oksanen, J. Olofsson, U. Rammul, M. Schneider, B.Johansen, and H. Olav Hygen. 2008. Spatial patterns anddynamic responses of arctic food webs corroborate theexploitation ecosystem hypothesis (EEH). American Natu-ralist 171:249–262.

Baermann, G. 1917. Eine eifache Methode Zur Auffindung vonAnklyostomum (Nematoden) larven in Erdproben. Geneesk.Tijdschrift voor Nederlands Indie 57:131–137.

Bardgett, R. D., and D. A. Wardle. 2010. Aboveground–belowground linkages: biotic interactions, ecosystem pro-cesses and global change. Oxford University Press, NewYork, New York, USA.

Batzli, G. O., and H. Henttonen. 1990. Demography andresource use by microtine rodents near Toolik Lake, Alaska.Arctic and Alpine Research 22:51–64.

Batzli, G. O., and C. Lesieutre. 1995. Community organizationof arvicoline rodents in northern Alaska. Oikos 72:88–98.

Batzli, G. O., R. G. White, S. F. MacLean, Jr., F. A. Pitelka,and B. D. Collier. 1980. The herbivore-based trophic system.Pages 335–410 in J. Brown, P. C. Miller, L. L. Tieszen, andF. L. Bunnell, editors. An arctic ecosystem: the coastaltundra at Barrow, Alaska. Dowden, Hutchinson, and Ross,Stroudsburg, Pennsylvania, USA.

Bret-Harte, M. S., M. C. Mack, G. R. Goldsmith, D. B. Sloan,J. DeMarco, G. R. Shaver, P. M. Ray, Z. Biesinger, and F. S.Chapin, III. 2008. Plant functional types do not predictbiomass responses to removal and fertilization in Alaskantussock tundra. Journal of Ecology 96:713–726.

Bret-Harte, M. S., G. R. Shaver, and F. S. Chapin, III. 2002.Primary and secondary stem growth in arctic shrubs:implications for community response to environmentalchange. Journal of Ecology 90:251–267.

Cebrian, J., J. B. Shurin, E. T. Borer, B. J. Cardinale, J. T.Ngai, M. D. Smith, and W. F. Fagan. 2009. Producernutritional quality controls ecosystem trophic structure.PLoS ONE 4(3):e4929.

Cheng, W., R. A. Virginia, S. F. Oberbauer, J. D. Tenhunen,C. T. Gillespie, and J. F. Reynolds. 1998. Soil nitrogen,microbial biomass, and respiration along an arctic to-posequence. Soil Science Society of America Journal62:654–662.

Clemmensen, K. E., A. Michelsen, S. Jonasson, and G. R.Shaver. 2006. Increased ectomycorrhizal fungal abundanceafter long-term fertilization and warming of two arctictundra ecosystems. New Phytologist 171:391–404.

De Deyn, G. B., C. E. Raaijmakers, and W. H. van der Putten.2004. Plant community development is affected by nutrientsand soil biota. Journal of Ecology 92:824–834.

Doles, J. 2000. A survey of soil biota in the arctic tundra andtheir role in mediating terrestrial nutrient cycling. MastersThesis. University of Northern Colorado, Greeley, Colorado,USA.

Fraser, L. H., and J. P. Grime. 1999. Experimental tests oftrophic dynamics: towards a more penetrating approach.Oecologia 119:281–284.

Gornall, J. L., S. J. Woodin, I. S. Jonsdottir, and R. van derWal. 2009. Herbivore impacts to the moss layer determinetundra ecosystem response to grazing and warming. Oeco-logia 161:747–758.

Gough, L., E. A. Ramsey, and D. R. Johnson. 2007. Plant–herbivore interactions in Alaskan arctic tundra change withsoil nutrient availability. Oikos 116:407–418.

Gough, L., K. Shrestha, D. R. Johnson, and B. Moon. 2008.Long-term mammalian herbivory and nutrient addition alterlichen community structure in Alaskan dry heath tundra.Arctic, Antarctic, and Alpine Research 40:65–73.

Gough, L., P. A. Wookey, and G. R. Shaver. 2002. Dry heatharctic tundra responses to long-term nutrient and lightmanipulation. Arctic, Antarctic, and Alpine Research34:211–218.

Graglia, E., R. Julkunen-Tiitto, G. R. Shaver, I. K. Schmidt, S.Jonasson, and A. Michelsen. 2001. Environmental controland intersite variations of phenolics in Betula nana in tundraecosystems. New Phytologist 151:227–236.

Grellmann, D. 2002. Plant responses to fertilization andexclusion of grazers on an arctic tundra heath. Oikos98:190–204.

Gruner, D. S., J. E. Smith, E. W. Seabloom, S. A. Sandin, J. T.Ngai, H. Hillebrand, W. S. Harpole, J. J. Elser, E. E. Cleland,M. E. S. Bracken, E. T. Borer, and B. M. Bolker. 2008. Across-system synthesis of consumer and nutrient resourcecontrol on producer biomass. Ecology Letters 11:740–755.

Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960.Community structure, population control, and competition.American Naturalist 94:421–425.

Hik, D. S., and R. L. Jefferies. 1990. Increases in the NAPP of asalt-marsh forage grass: a test of the predictions of theherbivore-optimization model. Journal of Ecology 78:180–195.

Hilbert, D. W., D. M. Swift, J. K. Detling, and M. I. Dyer.1981. Relative growth rates and the grazing optimizationhypothesis. Oecologia 51:14–18.

Hobbie, S. E. 1996. Temperature and plant species control overlitter decomposition in Alaskan tundra. Ecological Mono-graphs 66:503–522.

Hobbie, S. E., L. Gough, and G. R. Shaver. 2005. Speciescompositional differences on different-aged glacial land-scapes drive contrasting responses of tundra to nutrientaddition. Journal of Ecology 93:770–783.

Hunt, H. W., D. C. Coleman, E. R. Ingham, R. E. Ingham,E. T. Elliott, J. C. Moore, S. L. Rose, C. P. P. Reid, andC. R. Morley. 1987. The detrital food web in a shortgrassprairie. Biology and Fertility of Soil 3:57–68.

July 2012 1693ABOVE- AND BELOWGROUND TEST OF EEH

Johnson, D. R. 2008. How herbivores affect plant growth,community structure and litter decomposition in Alaskantundra: implications for responses to climate change.Dissertation. University of Texas, Arlington, Texas, USA.

Lenhart, E. A. 2002. Central Arctic Herd. Pages 296–292 in C.Brown, editor. Caribou Management Report of Survey ofInventory Activities. Alaska Department of Fish and Game,Division of Wildlife Conservation, Juneau, Alaska, USA.

Leriche, H., X. LeRoux, and J. Gignoux, A. Tuzet, H. Fritz, L.Abbadie, and M. Loreau. 2001. Which functional processescontrol the short-term effect of grazing on net primaryproductivity in grasslands? Oecologia 129:114–124.

MacArthur, R. H., and E. O. Wilson. 1967. The theory ofisland biogeography. Princeton University Press, Princeton,New Jersey, USA.

Moore, J. C., and P. C. de Ruiter. 2012. Energetic food webs:An analysis of real and model ecosystems. Oxford UniversityPress, Oxford, UK. In press.

Moore, J. C., P. C. de Ruiter, and H. W. Hunt. 1993. Influenceof productivity on the stability of real and model ecosystems.Science 261:906–908.

Moore, J. C., K. McCann, H. Setala and P. C. de Ruiter. 2003.Top-down is bottom-up: Does predation in the rhizosphereregulate aboveground production? Ecology 84:846–857.

Moore, J. C., B. B. Tripp, R. T. Simpson, and D. C. Coleman.2000. Springtails in the classroom: Collembola as modelorganisms for inquiry-based laboratories. American BiologyTeacher 62:512–519.

Moore, J. C., D. E. Walter, and H. W. Hunt. 1988. Arthropodregulation of micro- and mesobiota in belowground detritalfood webs. Annual Review of Entomology 33:419–439.

Odum, E. P. 1969. The strategy of ecosystem development.Science 164:262–279.

Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemel. 1981.Exploitative ecosystems in gradients of primary productivity.American Naturalist 118:240–261.

Olofsson, J., H. Kitti, P. Rautiainen, S. Stark, and L. Oksanen.2001. Effects of summer grazing by reindeer on compositionof vegetation, productivity and nitrogen cycling. Ecography24:13–24.

Olofsson, J., L. Oksanen, T. Callaghan, P. E. Hulme, T.Oksanen, and O. Suominen. 2009. Herbivores inhibitclimate-driven shrub expansion on the tundra. GlobalChange Biology 15:2681–2693.

Olofsson, J., S. Stark, and L. Oksanen. 2004. Reindeer influenceecosystem processes in the tundra. Oikos 105:386–396.

Pajunen, A., R. Virtanen, and H. Roininen. 2008. The effects ofreindeer grazing on the composition and species richness ofvegetation in forest–tundra ecotone. Polar Biology 31:1233–1244.

Post, E., et al. 2009. Ecological dynamics across the Arcticassociated with recent climate change. Science 325:1355–1358.

Post, E., and C. Pedersen. 2008. Opposing plant communityresponses to warming with and without herbivores. Proceed-ings of the National Academy of Sciences USA 105:12353–12358.

Schmitz, O. J. 2008. Herbivory from individuals to ecosystems.Annual Review of Ecology, Evolution, and Systematics39:133–152.

Schmitz, O. J., V. Krivan, and O. Ovadia. 2004. Trophiccascades: the primacy of trait-mediated indirect interactions.Ecology Letters 7:153–163.

Shaver, G. R., M. S. Bret-Harte, M. H. Jones, J. Johnstone, L.Gough, J. Laundre, and F. S. Chapin, III. 2001. Speciescomposition interacts with fertilizer to control long-termchange in tundra productivity. Ecology 82:3163–3181.

Stark, S., R. Strommer, and J. Tuomi. 2002. Reindeer grazingand soil microbial processes in two suboceanic and twosubcontinental tundra heaths. Oikos 97:96–78.

Strong, D. R. 1992. Are trophic cascades all wet? Differenti-ation and donor-control in speciose ecosystems. Ecology73:747–754.

Sullivan, P. F., M. Sommerkorn, H. M. Rueth, K. J.Nadelhoffer, G. R. Shaver, and J. M. Welker. 2007. Climateand species affect fine root production with long-termfertilization in acidic tussock tundra near Toolik Lake,Alaska. Oecologia 153:643–652.

van der Wal, R. 2006. Do herbivores cause habitat degradationor vegetation state transition? Evidence from the tundra.Oikos 114:177–186.

van der Wal, R., R. D. Bardgett, and K. A. Harrison. 2004.Vertebrate herbivores and ecosystem control: cascadingeffects of faeces on tundra ecosystems. Ecography 27:242–252.

van der Wal, R., and R. W. Brooker. 2004. Mosses mediategrazer impacts on grass abundance in arctic ecosystems.Functional Ecology 18:77–86.

van Wijk, M. T., et al. 2003. Long-term ecosystem levelexperiments at Toolik Lake, Alaska, and at Abisko,Northern Sweden: generalizations and differences in ecosys-tem and plant type responses to global change. GlobalChange Biology 10:105–123.

Wardle, D. A. 2002. Communities and ecosystems: linking theaboveground and belowground components. Monographs inPopulation Biology. Princeton University Press, Princeton,New Jersey, USA.

Wookey, P. A., R. Aerts, R. D. Bardgett, F. Baptist, K. A.Brathen, J. H. C. Cornelissen, L. Gough, I. P. Hartley, D. W.Hopkins, S. Lavorel, and G. R. Shaver. 2009. Ecosystemfeedbacks and cascade processes: understanding their role inthe responses of Arctic and alpine ecosystems to environ-mental change. Global Change Biology 15:1153–1172.

Yeates, G. W., T. Bongers, R. G. M. De Goede, D. W.Freckman, and S. S. Georgieva. 1993. Feeding habits in soilnematode families and genera: an outline for soil ecologists.Journal of Nematology 25:315–331.

LAURA GOUGH ET AL.1694 Ecology, Vol. 93, No. 7