|
Received: 21 June 2016    Accepted: 27 June 2017
DOI: 10.1111/1365-2656.12744
RESEARCH ARTICLE
Mammal-induced trophic cascades in invertebrate food webs
are modulated by grazing intensity in subalpine grassland
Martijn L. Vandegehuchte1,2  
1
Research Unit Community Ecology, Swiss
Federal Institute for Forest, Snow and
Landscape Research, Birmensdorf, Switzerland
2
Terrestrial Ecology Unit, Department of
Biology, Ghent University, Ghent, Belgium
3 apex predators and their top-down forces. Bottom-up trophic interaction chains
PXL-Tech, PXL University College, Hasselt,
Belgium
4
Institute of Environmental
Engineering, Department of Civil,
Environmental and Geomatic
Engineering, ETH Zurich, Zurich, Switzerland
Correspondence
Martijn L. Vandegehuchte
Email: martijn.vandegehuchte@ugent.be
Funding information
Swiss National Science Foundation, Grant/
Award Number: 31003A_122009/1 and
31003A_140939/1
Handling Editor: Peter Hambäck
1434  |  © 2017 The Authors. Journal of Animal Ecology
© 2017 British Ecological Society
| Martin Schütz1 | Frederic de Schaetzen3,4 | Anita C. Risch1
Abstract
1. Even though mammalian herbivores can exert strong indirect effects on other ani-
mals by altering the vegetation, the study of trophic cascades retains a focus on
induced by mammalian herbivores, particularly in invertebrate food webs, remain
largely unexplored.
2. We tested whether effects of mammalian herbivores on the vegetation ricochet
back up several trophic levels of the invertebrate food web. We further tested two
alternative hypotheses: the strength of herbivore-induced indirect interactions ei-
ther increases with plant productivity because of a concurrent higher grazing inten-
sity, or it decreases because of a higher plant tolerance to grazing.
3. We progressively excluded large, medium and small herbivorous mammals from
replicated plots of 6 m2 in productive, intensively grazed short-grass vegetation
and less productive, less intensively grazed tall-grass vegetation of subalpine grass-
lands. We measured vegetation quantity, quality, structure and composition, and
determined the abundance of invertebrate herbivores, detritivores, omnivores and
predators. We used structural equation modelling to test vegetation-mediated cas-
cading effects of the different mammalian herbivores across different trophic
groups of invertebrates.
4. In the short-grass vegetation, mammals caused changes in vegetation quantity and
thickness. These changes directly affected detritivorous and predatory inverte-
brate abundance, yet indirectly affected predatory and omnivorous invertebrates
through a bottom-up trophic cascade via changes in herbivorous invertebrate
abundance. In the tall-grass vegetation, mammal-induced changes in vegetation
quality and composition affected detritivorous invertebrates and in turn omnivo-
rous invertebrates, but these cascading effects were weaker than those in the
short-grass vegetation. Smaller mammals were at least as important as large mam-
mals in structuring the invertebrate food web.
5. Our results demonstrate that differently sized mammalian herbivores can trigger
trophic cascades in the local invertebrate food web. Our findings further support
the hypothesis that herbivore-induced indirect interactions are stronger in more
productive systems because of higher foraging intensity, as opposed to the hypoth-
esis that a higher grazing tolerance of plants should dampen herbivore-induced
indirect interactions in productive systems.
wileyonlinelibrary.com/journal/jane J Anim Ecol. 2017;86:1434–1446.
VANDEGEHUCHTE ET AL.
KEYWORDS
competition, herbivory, insects, interaction chains, plants, rodents, ungulates
1  |  INTRODUCTION
More than half a century ago, Hairston, Smith, and Slobodkin (1960) fa-
mously proposed that the “world is green” because herbivores are kept
in check through top-down control by predators. What followed were
decades of debate on the role of bottom-up and top-down forces and
the prevalence of trophic cascades in food webs. The initial dichotomy
between bottom-up (White, 1978) or top-down (Hairston et al., 1960;
Oksanen, Fretwell, Arruda, & Niemelä, 1981) perspectives has largely
been replaced by the consensus view that most systems are organized
according to a “bottom-up template” onto which top-down effects can
be superimposed (Hunter & Price, 1992). Despite this consensus, the
study of “cascading” effects in ecological interaction webs has retained
a predominantly top-down focus (but see Kagata & Ohgushi, 2006;
Pringle, Young, Rubenstein, & McCauley, 2007), with an emphasis
on the multitrophic effects of top predators (Crooks & Soule, 1999;
Terborgh et al., 2001) and the determinants of predator-induced tro-
phic cascades across ecosystems (Borer et al., 2005; Mooney et al.,
2010; Schmitz, Hambäck, & Beckerman, 2000; Shurin et al., 2002).
Theory about bottom-up vs. top-down effects and trophic cascades
largely deals with controls on population size (Hairston et al., 1960;
Hunter & Price, 1992; Oksanen et al., 1981; White, 1978). In contrast,
empirical evidence is biased towards experimental scales smaller than
the scale of resource or consumer populations (e.g. Borer et al., 2005;
Kagata & Ohgushi, 2006; Mooney et al., 2010). However, spatial scale
can affect the interpretation of trophic interaction experiments. For
example, predator effects are often tested using cages permeable to
prey. As the size of such cages decreases, predation effects become in-
creasingly driven by prey movement relative to prey vital rates, and can
differ from true population effects (Englund, 1997). Similarly, effects
of experimental herbivore exclusions on plants have been shown to be
scale-dependent (Gil, Jiao, & Osenberg, 2016).
Even though mammalian herbivores are expected to exert “ram-
pant indirect effects” within food webs (Paine, 2000), they have re-
ceived much less attention as potential initiators of multitrophic
interaction chains than large predators (Pringle et al., 2007). This was
emphasized by Foster, Barton, and Lindenmayer (2014) in a recent
meta-analysis, highlighting a major gap in knowledge on the mecha-
nisms by which large mammalian herbivores have cascading effects
across the rest of the food web. A few studies have shown that grazing
ungulates induce bottom-up trophic cascades involving invertebrates
by altering vegetation properties (Evans et al., 2015; Nuttle, Yerger,
Stoleson, & Ristau, 2011; Pringle et al., 2007; Vanbergen, Hails, Watt,
& Jones, 2006), but cascading effects of mammalian herbivores across
entire invertebrate trophic levels remain untested. Moreover, the con-
tributions of smaller herbivorous mammals (e.g. rodents, lagomorphs)
to such cascades are unexplored, even though the importance of
small mammals for ecosystem properties is increasingly recognized
Journal of Animal Ecology |
      1435
(e.g. Bakker, Ritchie, Olff, Milchunas, & Knops, 2006; Howe, Brown,
& Zorn-Arnold, 2002; Howe, Zorn-Arnold, Sullivan, & Brown, 2006;
Rebollo, Milchunas, Stapp, Augustine, & Derner, 2013).
Pathways by which mammalian herbivores affect the vegetation
and therefore invertebrates are manifold. The most immediate trophic
effect of herbivorous mammals on herbivorous and detritivorous in-
vertebrates is lowering the availability of food resources by removing
plant biomass and decreasing the amount of leaf litter (Ford, Garbutt,
Jones, & Jones, 2013; Gómez & González-Megías, 2002; Huntly, 1991;
Takagi & Miyashita, 2014). Mammalian herbivores can also cause shifts
in plant species composition through selective feeding (e.g. Bardgett &
Wardle, 2003; Howe et al., 2002; Milchunas & Lauenroth, 1993) and
alter the nutritional quality of plants and leaf litter (McNaughton, 1985;
Ritchie, Tilman, & Knops, 1998; Wardle, Bonner, & Barker, 2002).
Mammalian herbivores can furthermore affect plant growth and qual-
ity through the redistribution of nutrients via dung and urine (Bardgett
& Wardle, 2003; Ritchie et al., 1998). In productive systems with high
mammalian grazing pressure, grazers tend to benefit grazing-adapted,
nutrient-rich plant species, whereas in low-productivity systems, graz-
ers tend to disadvantage palatable plant species, leading to the re-
placement of palatable plants by better-defended plants with lower
foliage and litter quality (Bardgett & Wardle, 2003). These changes in
vegetation quality can have trophic effects on herbivorous and detri-
tivorous invertebrates (Danell & Hussdanell, 1985; Takagi & Miyashita,
2012). Similarly, changes in vegetation composition can alter resource
heterogeneity, which can also affect the abundance of herbivorous in-
vertebrates (Loranger et al., 2014; Scherber et al., 2010) or, if the com-
position of leaf litter is affected, detritivorous invertebrate abundance
(Moore et al., 2004; Wardle, Yeates, Barker, & Bonner, 2006). Each of
these vegetation-mediated effects of mammalian herbivores on inver-
tebrate herbivores and detritivores can indirectly influence predatory
and omnivorous invertebrates feeding on these lower trophic level
invertebrates. In addition, as “ecosystem engineers” (Jones, Lawton,
& Shachak, 1997; Prugh & Brashares, 2012), grazing mammals can
have non-trophic effects on invertebrates: by removing plants or plant
parts, or indirectly by altering the species composition of the vegeta-
tion, they can alter the physical habitat of invertebrates of all feeding
types (Milchunas & Lauenroth, 1993; Rebollo et al., 2013).
The strength of the above-mentioned effects of mammalian her-
bivores on invertebrates has been suggested to depend on plant pro-
ductivity, according to two contrasting hypotheses (Daskin & Pringle,
2016). One hypothesis is based on the premise that low-productivity
environments have low plant biomass and select for well-defended, un-
palatable plants. Therefore, herbivores should be more abundant and
graze at a higher intensity in high-productivity systems. Mammalian
herbivore effects on the vegetation, and consequently on invertebrate
communities, are hence predicted to increase in strength with plant
productivity (“foraging-intensity” hypothesis; Daskin & Pringle, 2016).
 |
1436      
Journal of Animal Ecology
The alternative hypothesis posits that higher productivity increases
the ability of plants to regrow and tolerate herbivory and that in the
most productive systems herbivores can only consume a small frac-
tion of total biomass. Therefore, the strength of mammalian herbivore
effects on plants and invertebrates should decrease with plant pro-
ductivity (“tolerance/avoidance” hypothesis; Daskin & Pringle, 2016).
To test these hypotheses, we performed an extensive field exper-
iment in which we progressively excluded large (ungulates), medium-
sized (marmots, hares) and small (mice, voles) mammals in two
vegetation types that together cover the entire subalpine grasslands in
our study area: productive short- and less productive tall-grass vegeta-
tion. Both bottom-up and top-down forces can operate in invertebrate
food webs (Gruner, 2004; Rosenheim, 1998), but bottom-up mecha-
nisms seem to prevail in terrestrial invertebrate communities (Gruner,
2004; Rzanny, Kuu, & Voigt, 2013; Scherber et al., 2010). Changes in
vegetation properties caused by mammalian herbivores are therefore
expected to result in bottom-up cascades in invertebrate food webs.
Our aim was to test the contrasting “foraging-intensity” and “toler-
ance/avoidance” hypotheses regarding the relationship between the
strength of such mammal-induced bottom-up cascades and ecosystem
productivity. Mammalian herbivores exert a higher grazing pressure in
the short- than in the tall-grass vegetation (Risch, Haynes, Busse, Filli,
& Schütz, 2013). Therefore, we predict the interaction chains between
mammal exclusions, vegetation properties and invertebrate trophic
groups to be more numerous and composed of stronger links in the
more productive and more intensively grazed short-grass vegetation,
in accordance with the “foraging-intensity” hypothesis.
The small size of our mammal exclosures (6 m2) implies that each
of the above-mentioned vegetation-mediated effects of mammals on
invertebrates could operate via changes in the invertebrates’ migra-
tion into or out of experimental plots as well as via changes in their
vital rates (reproduction, mortality). For example, if the quantity or
quality of the local vegetation increases, herbivorous and detritivorous
invertebrates may increase local reproduction and survival as well as
migration into the plot. This could in turn increase the immigration
as well as the local performance of omnivorous and predatory inver-
tebrates. As the relative importance of movement vs. vital rates in-
creases at smaller spatial scales (cf. Englund, 1997), we caution against
extrapolating our results to the scale of an entire grassland and its
invertebrate populations.
2  |  MATERIALS AND METHODS
2.1 | Experimental design
The experiment has been described in detail in Risch et al. (2013) and
Vandegehuchte, Raschein, Schütz, Gwiazdowicz, and Risch (2015). In
short, six subalpine grasslands were selected in the Swiss National
Park (SNP; south-eastern Switzerland). Each grassland is entirely made
up of two co-occurring vegetation types resulting from long-term dif-
ferences in land-use history and ungulate grazing intensity. Short-
grass vegetation, dominated by lawn grasses, is found where cattle
and sheep rested for over five centuries. The animals transferred high
VANDEGEHUCHTE ET AL.
levels of nutrients to the soils of these areas. Livestock was banned
when the SNP was established in 1914. Since then, naturally re-
immigrated wild ungulates have been preferentially feeding in these
high-nutrient areas, turning the vegetation into short-grass grazing
lawns (Schütz et al., 2006). Tall-grass vegetation is dominated by tus-
sock graminoids and occurs where livestock grazed, but did not rest,
prior to park establishment. Since the foundation of the park, this
vegetation type has been grazed much less intensively by wild ungu-
lates than the short-grass vegetation due to its lower nutritional value.
During our 5-year experiment (2009–2013), roughly one hundred
years after the park’s foundation, the short-grass vegetation was still
consistently more productive than the tall-grass vegetation (583 vs.
523 g dry biomass m−2 year−1) and much more intensively grazed (316
vs. 99 g dry biomass m−2 year−1 consumed), resulting in lower stand-
ing plant biomass (267 vs. 424 g dry biomass m−2 year−1). The mam-
mals inhabiting these grasslands can be categorized according to their
body size: large (red deer [Cervus elaphus L.] and chamois [Rupicapra
rupicapra L.], 30–150 kg), medium-sized (alpine marmot [Marmota
marmota L.] and mountain hare [Lepus timidus L.], 3–6 kg) and small
(small rodents, e.g. Clethrionomys spp., Microtus spp., Apodemus spp.,
30–100 g).
We erected two exclosure set-ups per vegetation type in the three
large grasslands, and one per vegetation type in the three smaller
grasslands (18 exclosure set-ups in total). One set-up (Figure 1) con-
sisted of five 2 × 3 m plots, of which one was unfenced so that all an-
imals had access (“Control”, i.e. none of the mammals excluded). The
control plots were located at least 5 m away from the main electrical
fence, which enclosed the remaining four plots and excluded ungulates
(for details, see Risch et al., 2013). Within the main fence, one plot was
left unfenced, so that except ungulates, all other animals had access
(“Deer exclosure”). A second plot was fenced with electric netting to
further exclude medium-sized mammals (“Deer-marmot exclosure”).
The lowest openings in this netting were not electrified, to ensure
that small mammals would not be deterred from entering. A third plot
was fenced with metal wire netting to also exclude small mammals
(“Deer-marmot-mouse exclosure”). The fourth plot was enclosed by
a mosquito-netting fence and roof to exclude all above-ground verte-
brates and invertebrates, but no data from this treatment were used
in this study. Consequently, data obtained from 72 treatment plots
were used (18 exclosure set-ups × 4 treatment plots). Each 2 × 3 m
treatment plot was divided into six 1 × 1 m subplots used to assess
different ecosystem properties (Figure 1). Exclosure set-ups were first
erected in the spring of 2009. Each year they were removed in the fall
(late October) to prevent damage from snow pressure and avalanches
and reinstalled the next spring (early May), immediately after snow-
melt. The exclosure set-ups were in place for five consecutive grow-
ing seasons (2009–2013). Ungulates migrate to lower altitudes and
marmots hibernate during winter, yet small rodents may have grazed
on vegetation under the snow when the exclosure set-ups where not
in place. However, we did not observe any signs of extensive winter
activity by small mammals after snowmelt (e.g. extensive burrows).
Treatment plots had to be kept rather small (2 × 3 m), because of strict
limitations on the size of research infrastructure allowed in the SNP.
VANDEGEHUCHTE ET AL.
F I G U R E 1   Schematic overview of one of the 18 exclosure
set-ups. The main outer fence excluding ungulates as well as the
fence surrounding the 2 × 3 m plot to further exclude marmots and
hares were connected to a solar-charged power source. Each of the
different exclusion treatments was assigned at random to one of the
four 2 × 3 m plots within the outer fence. Subplots 1, 5 and 6 were
not used in this study. Two pitfall traps were installed in the inner
corners of subplot 2. Suction sampling also took place in subplot
2. Subplot 3 was used for estimation of plant species cover and for
parallel photography of the vegetation. Vegetation was clipped in
subplot 4 for analysis of nitrogen content. See main text for more
details
2.2 | Invertebrate sampling
Invertebrates were sampled during the 2013 growing season, that is
only during the fifth year of the exclusion experiment. Plant-dwelling
invertebrates were collected monthly from June to September 2013
(four times) using a suction sampler (Vortis, Burkhard Manufacturing
Co. Ltd.). A square plastic frame (60 × 60 × 40 cm) with closable
mosquito mesh sleeve on top was quickly placed into the treatment
plot (subplot 2, Figure 1). The suction sampler was then inserted
Journal of Animal Ecology |
      1437
through the sleeve and operated for 45 s to sample any enclosed in-
vertebrates. Sample vials were immediately filled with 70% ethanol.
Ground-dwelling invertebrates were sampled using pitfall traps. Two
traps were installed in each treatment plot at the end of May 2013
(subplot 2, Figure 1). Each pitfall trap consisted of a plastic cylinder
(13 cm length, 6.75 cm inner diameter), placed into the soil, flush with
the soil surface. We placed a 100-ml vial (6.70 cm outer diameter)
containing 20% propylene glycol into the cylinder. A plastic funnel
was suspended in the top part of the cylinder to guide invertebrates
into the vial. A transparent plastic cone-shaped roof was placed over
each pitfall trap to shield the trap from rain. Pitfall traps were emptied
biweekly until early September (seven times). Contents were rinsed
with tap water and stored in 70% ethanol. Invertebrates from pitfall
and suction samples were sorted into higher-order taxonomic groups
in the laboratory. Each invertebrate taxon was classified into one of
the following feeding types: herbivore, detritivore, omnivore, preda-
tor, nectar feeder, mixed (taxon containing several feeding types),
not feeding as adult or unknown (see Table S1). For this study, we
only used data pertaining to herbivores, detritivores, omnivores and
predators.
2.3 | Vegetation characteristics
The vegetation structure was assessed in July 2013 using parallel pho-
tography followed by digital image analysis (see Figure S1). In each
treatment plot, a strip of vegetation (100 cm long, 30 cm wide) was
selected (subplot 3, Figure 1). Behind this strip, we placed a 130 cm
wide and 60 cm high matte blue canvas with a black 100 cm scale.
Vegetation in front of the selected area was covered with a mulch-
ing fleece. A digital single-lens reflex camera (Nikon D60 + Nikon
18–55 mm f/3.5-5.6G AF-S VR DX NIKKOR, Nikon) was placed
140 cm from the background canvas and 25 cm above the soil surface.
A photograph in Nikon RAW format (.NEF) was taken at a focal length
of 18 mm. Sharpness and contrast were enhanced in Adobe Lightroom
4, and the photograph was cropped (2:1.2 aspect ratio) so that the long
edge of the photograph coincided with the 100-cm scale on the back-
ground canvas. The photograph was then exported in JPEG format at
2,000 × 1,200 pixels, corresponding to 100 × 60 cm (0.6 m2) in real
life, and converted to a black-and-white 8-bit file using the thresh-
old colour function in IMAGEJ (National Institutes of Health). This file
was processed in SideLook (Zehm, Nobis, & Schwabe, 2003) to assess
several vegetation structure properties. The number of black pixels
was divided by the total number of pixels (2,000 × 1,200 = 2,400,000)
and multiplied by 0.6 to calculate the vegetation quantity in m2/m,
that is the total area of black pixels in relation to the picture width
of 1 m. The height of the vegetation (cm) for each column of 1 cm
width was determined as the height of the highest black pixel in that
column. This series of height values was subsequently averaged. We
then calculated the ratio of vegetation quantity to average height, as a
measure of vegetation thickness, that is the amount of vegetation per
unit height (m2 m−1 cm−1), with higher and lower values, respectively,
representing denser and sparser vegetation. Plant species composi-
tion was assessed by visually estimating percentage cover for each
 |
1438      
Journal of Animal Ecology
plant species at peak biomass (July 2013) in a 1 × 1 m subplot of each
treatment plot (subplot 3, Figure 1). Plant biomass clipped from two
10 × 100 cm strips in another 1 × 1 m subplot of each treatment plot
(subplot 4, Figure 1) was combined, dried, ground (Pulverisette 16,
Fritsch) and passed through a 0.5-mm sieve. Nitrogen (N) concentra-
tion was determined as described in Vandegehuchte et al. (2015).
2.4 | Statistical analyses
We first summed the abundance of each invertebrate taxon across
the two pitfall traps in each treatment plot at each sampling time, to
obtain a single value per plot and time. Despite the roofs a few pitfall
traps were flooded (run-off). If only one of the two traps per plot was
flooded, we doubled the abundance of the non-flooded pitfall trap for
that sampling time. If both pitfall traps were flooded, we assigned the
average of the summed values of the previous and following sampling
time to this plot. Subsequently, we summed the abundances of her-
bivores, detritivores, omnivores and predators across the seven pitfall
emptying events. We similarly summed the abundances of each of
these trophic groups across the four suction sampling events. Juvenile
leafhoppers, juvenile spiders and springtails were omitted from analy-
ses because they showed some very large outliers, caused for exam-
ple by hatching of leafhopper egg clusters, female spiders carrying
nymphs on the back and soil-dwelling springtails being sucked out of
the soils. For one plot, the value for omnivore abundance was omitted
from the suction sampling data because this plot contained a hidden
ant nest from which hundreds of ants were sucked into the sampling
vial, resulting in a very large outlier. Total numbers of individuals per
sampling method and taxonomic group are given in Table S1.
Structural equation modelling (SEM) is a preferred technique for
analysing complex networks of relationships and can identify support
for causal linkages that bivariate analyses fail to clarify (Grace, 2006).
We therefore used SEM to test our multivariate hypotheses about
plant-mediated effects of mammals on invertebrate food webs. Our
hypothetical model included the effects of each of the different mam-
mal exclusion treatments on each of four vegetation characteristics:
vegetation quality (N content), vegetation quantity, vegetation thick-
ness and vegetation composition (Principal Coordinate Analysis [PCO]
axis 1 scores based on Bray–Curtis dissimilarities in plant species
cover between samples). The exclusion treatments were modelled as
dummy variables, with the “Control” plots acting as reference level, as
is usual for categorical exogenous variables in SEM (Grace, 2006). The
model included residual covariances between each pair of vegetation
characteristics, which represent unexplained relationships between
these variables. Each vegetation characteristic was hypothesized to
influence the abundance of invertebrate herbivores, omnivores and
detritivores, as these groups potentially depend on the quality, quan-
tity, structure and composition of the vegetation. As predators do not
directly consume plant biomass, we did not include a link from plant
quality to predator abundance. Links from vegetation quantity, thick-
ness and composition to predator abundance were, however, included,
as these can influence predatory invertebrates by altering habitat
structure and microclimate. We expected that vegetation-mediated
VANDEGEHUCHTE ET AL.
effects of mammal exclusions on lower trophic levels (herbivores,
detritivores) would cascade up the food web to affect higher trophic
levels (omnivores, predators) in a bottom-up fashion (see Section 1).
We therefore included effects of herbivores and detritivores on pred-
ators and omnivores, and from omnivores on predators. We realize
that some of these bottom-up trophic links could run in the opposite
direction, that is top-down. However, the recommended strategy is to
test the hypothetical model with the highest level of theoretical sup-
port, that is our bottom-up model (Grace, 2006). If the overall model fit
was good, we accepted the multivariate hypothesis represented by the
model. If the fit was poor, we assessed whether a minimum of theoret-
ically meaningful path additions would improve the model. We did not
include direct effects of mammal exclusions on the different inverte-
brate groups in our initial model, as we expected accidental ingestion
or direct disturbance of invertebrates to be negligible in our system.
We also did not explore nonlinear relationships, as we expected ef-
fects to be predominantly linear in our system, and because the bio-
logical meaning of nonlinear relationships can be difficult to interpret
in complex SEMs.
To assess exclusion treatment effects independently from varia-
tion between exclosure set-ups, we calculated residuals per exclosure
set-up, scaled them to a constant variance of 1 and used these scaled
residuals in the SEM. A separate SEM was tested for each combination
of vegetation type and invertebrate sampling method. We decided to
analyse data from suction samples and pitfall traps separately, because
these methods target vegetation- and ground-dwelling invertebrates,
respectively, and differ in sampling efficiency. Because the suction
samples contained almost no detritivorous invertebrates (two individ-
uals in total), detritivores were omitted from the SEMs based on suc-
tion sampling data. Multivariate normality of the data was tested using
Henze–Zirkler’s and Mardia’s multivariate normality tests. Maximum
likelihood estimation was used to fit the data to the model by com-
paring the model-implied variance–covariance matrix against the
observed variance–covariance matrix. Suction sample and pitfall trap
data from the short-grass vegetation were not multivariate normal ac-
cording to both tests. Pitfall trap data from the tall-grass vegetation
were multivariate normal according to both tests, whereas suction
sampling data from this vegetation type were multivariate normal ac-
cording to Henze–Zirkler’s, but not to Mardia’s test. If one of the tests
indicated non-normality, we used maximum likelihood estimation with
“Huber–White” robust standard errors and a robust χ2 test statistic
that is asymptotically equivalent to the Yuan–Bentler T2-star test sta-
tistic. A large p-value (>.05) associated with the χ2 test statistic indi-
cates that the covariance structure of the data does not significantly
differ from the one expected based on the model (Grace, 2006). If
the p-value was smaller than .05, we investigated modification indices
(which provide an indication of the improvement in fit if a certain path
is added to the model) and correlation residuals (which indicate model
misspecification), and added the regression path with the largest mod-
ification index whose correlation residual had an absolute value larger
than 0.1 (Kline, 2005). Only paths that were theoretically plausible
were considered, including top-down paths from higher to lower tro-
phic level invertebrates and direct effects of mammal exclusions on
VANDEGEHUCHTE ET AL. Journal of Animal Ecology |
      1439

the different invertebrate groups. This model-building procedure was
repeated until model fit was adequate.
In addition to SEM, we also tested the overall effect of vegeta-
tion type and exclusion treatments on the individual vegetation and
invertebrate variables (the latter per sampling method), analysing the
experiment as a split-plot design with vegetation type as whole-plot
factor, exclusion treatment as split-plot factor and exclosure set-ups
(whole plots) blocked on grassland. We used general or generalized
linear mixed models with vegetation type, exclusion treatment and
their interaction as fixed effects, and exclosure set-up and grassland
as random effects. A Gaussian distribution was used for vegetation
thickness and composition (as their conditional studentized residuals
were normally distributed based on visual inspection and a Shapiro–
Wilk test), a Gamma distribution for vegetation quantity and quality,
and a negative binomial distribution and a log link function for the
abundance of each of the invertebrate groups. The false discovery rate
was used to correct post hoc pairwise comparisons for multiple testing
(Benjamini & Hochberg, 1995).
3  |  RESULTS
The SEM for the pitfall trap data from the short-grass vegetation
(Figure 2a) obtained adequate fit after adding some direct pathways
from exclusion treatments to invertebrate feeding type abundances
(after Yuan–Bentler correction: χ2 = 14.16, df = 9, p = .12). The SEM
for the suction sampling data from this vegetation type (Figure 2b) fit
adequately after adding one direct pathway from the “Deer-marmot

(a)

(b)

F I G U R E 2   Result of the SEMs for both sampling methods in the short-grass vegetation. (a) SEM for pitfall catches. (b) SEM for suction
samples. The four levels of mammal exclusion treatment are modelled as three dummy variables that are compared against the control plots
to which all mammals had access. Rectangles represent variables included in the model. Single-headed arrows represent directional effects.
Double-headed arrows represent residual covariances, which indicate an unexplained relationship between two variables. Solid arrows:
significant paths (p < .05), dashed arrows: non-significant paths. Black arrows: positive relationships, red arrows: negative relationships. Values
associated with arrows are standardized path coefficients, and arrow thickness is proportional to this standardized path coefficient. *p < .05,
**p < .01, ***p < .001. r2 values associated with endogenous variables indicate proportion of variation explained by relationships with other
variables. SEM, structural equation model; PCO, principal coordinate analysis
 |
1440      
Journal of Animal Ecology
exclosure” to predatory invertebrate abundance (after Yuan–Bentler
2
correction: χ = 15.99, df = 9, p = .067). The initial SEM based on pit-
fall trap data from the tall-grass vegetation (Figure 3a) showed an
adequate fit (χ2 = 17.42, df = 14, p = .23), whereas the SEM for the
suction sampling data from this vegetation type (Figure 3b) fit after
adding two additional direct pathways from exclusions to inverte-
brate abundances (after Yuan–Bentler correction: χ2 = 14.94, df = 8,
p = .060). Note that our model-building procedure never resulted in
top-down pathways from higher to lower trophic level invertebrates.
Variance–covariance matrices for short-grass SEMs are given in
Table S2 and for tall-grass SEMs in Table S3.
In the short-grass vegetation, the progressive exclusion of mam-
mal groups caused progressively increasing quantities of vegetation
(Figure 2a,b). Increases in vegetation quantity in turn showed a strong
positive link with the abundance of invertebrate herbivores, which
was further linked to increased invertebrate predator abundance
(Figure 2a,b). This bottom-up trophic cascade was very similar for both
pitfall trap and suction sampling data (Figure 2a,b). Ground-dwelling
omnivorous invertebrates also responded positively to the increase
(a)
(b)
F I G U R E 3   Result of the SEMs for both sampling methods in the tall-grass vegetation. (a) SEM for pitfall catches. (b) SEM for suction samples.
Exclusion treatment variables, coefficients, significance levels, r2 values and arrow shape, colour and thickness as in Figure 2. SEM, structural
equation model
VANDEGEHUCHTE ET AL.
in herbivorous invertebrates (Figure 2a). When all mammals were ex-
cluded (“Deer-marmot-mouse” exclosures), the vegetation thickness
was higher than in the control plots (Figure 2a,b). The increases in
vegetation quantity and thickness were related to decreased abun-
dances of detritivorous and predatory invertebrates caught in pitfall
traps (Figure 2a). The vegetation in the “Deer-marmot” exclosures
had a lower N content compared to the control plots, but vegetation
N content did not significantly relate to the abundance of any inver-
tebrate trophic group (Figure 2a,b). Vegetation composition did not
significantly differ between our mammal exclusion treatments, but
significantly related to the abundance of herbivores caught by suction
sampling, detritivores and omnivores (Figure 2a,b). The models for the
short-grass vegetation showed significant direct pathways from some
of the mammal exclusion treatments to certain invertebrate feeding
type abundances, indicating mechanisms that were not captured by
the vegetation characteristics we measured.
In the tall-grass vegetation, vegetation quantity significantly in-
creased compared to the control when ungulates and medium-sized
mammals were excluded, and even more strongly when all mammals
VANDEGEHUCHTE ET AL. Journal of Animal Ecology|
      1441

were excluded (Figure 3a,b). However, unlike in the short-grass veg-
etation, vegetation quantity was not significantly linked to any of the
invertebrate feeding types (Figure 3a,b). Vegetation thickness was unaf-
fected by the mammal exclusions and did not significantly relate to any
of the invertebrate feeding types (Figure 3a,b). Contrary to the short-
grass vegetation, the “Deer-marmot” exclosures had a higher vegetation
N content than the control plots, which positively linked to detritivore
abundance, and in turn omnivore abundance. However, this trophic cas-
cade of positive interactions was partly offset by changes in vegetation
composition, induced by the exclusion of large plus medium or all mam-
mals, which negatively related to detritivore abundance (Figure 3a).
These same changes in vegetation composition, however, positively
related to herbivorous invertebrates caught in pitfall traps (Figure 3a).
Averages per vegetation type and exclusion treatment of the dif-
ferent vegetation characteristics (Figure 4) and abundances of inver-
tebrate feeding types from pitfall traps and suction samples (Figure 5)
matched the combined direct and indirect effects of exclusions in
the SEM for each vegetation type and sampling method. A summary
of the statistical analyses of the overall effects of vegetation type
and mammal exclusion treatments on these variables is provided in
Table 1. These results show how SEM can reveal complex interactions
that cannot be deduced from investigating bivariate relationships be-
tween variables. As just one example, the abundance of detritivores
in the tall-grass vegetation was very similar across all four treatments
(Figure 5g). However, SEM revealed that this resulted from a positive
pathway from mammal exclusions to detritivores via changes in vege-
tation quality and an opposing negative link via changes in vegetation
composition (Figure 3a).
In summary, in the short-grass vegetation, more and tighter link-
ages between mammal exclusions, vegetation characteristics and the
different invertebrate trophic levels were observed. Conversely, in the
tall-grass vegetation, links between mammal exclusions and vegeta-
tion, and between vegetation and invertebrates were much more inde-
pendent. In both vegetation types, the strongest effects on vegetation
characteristics were observed when, in addition to ungulates, smaller
mammalian herbivores were excluded as well.
4  |  DISCUSSION
We tested whether mammalian herbivores of different body size
caused bottom-up cascading interactions across the invertebrate
food web by influencing the vegetation. We furthermore assessed
whether the strength of these interactions depended on the pro-
ductivity of the system. We found evidence for interactions be-
tween mammalian herbivores and single invertebrate feeding
groups, mediated solely by changes in vegetation properties, as
shown in several other studies (e.g. Danell & Hussdanell, 1985;
Huntzinger, Karban, & Cushman, 2008; Takagi & Miyashita, 2012).
However, top-down effects of mammalian herbivores on vegetation

(a) (b)
B
Vegetation quantity (m2/m)

1.85 0.20
Vegetation N content (%)

1.80 B
0.15 AB
1.75
A
1.70

1.65 0.10

1.60
0.05
1.55
Vegetation thickness (m2 m–1 cm–1)

(c) (d)
a
0.006 ab
Plant PCO axis 1 score

0.15
ab ab
0.10
ab
0.05
0.005
bc
0.00
–0.05 cd
cd
0.004 –0.10
–0.15

No. De. De. De. No. De. De. De.

Ma. Ma. Ma. Ma.
Mo. Mo.
Mammals excluded Mammals excluded

F I G U R E 4   Vegetation characteristics for the different mammal exclusion treatments in short- and tall-grass vegetation. (a) Vegetation
N content (%). (b) Vegetation quantity (m2/m). (c) Vegetation thickness (m2 m−1 cm−1). (d) Plant PCO axis 1 score. White circles: short-grass
vegetation. Black circles: tall-grass vegetation. M ± Cousineau–Morey SE (wide cap—normalized per exclosure set-up for treatment comparisons;
narrow cap—normalized per grassland for vegetation type comparisons, see Cousineau, 2005; Morey, 2008). Mammals excluded: No. = none;
De. = deer and chamois; Ma. = marmots and hares; Mo. = mice and voles. Different capital letters indicate significant main differences between
treatments after false discovery rate adjustment. Different small letters indicate significant differences between combinations of vegetation type
and treatment after false discovery rate adjustment. PCO, principal coordinate analysis
1442      
Journal of Animal Ecology | VANDEGEHUCHTE ET AL.

properties additionally caused bottom-up cascades linking several
invertebrate trophic levels, which to our knowledge has not been
documented before. These mammal-induced trophic cascades were
stronger in the more productive, intensively grazed short-grass
vegetation than in the less productive and less intensively grazed
tall-grass vegetation, in support of the “foraging-intensity” hypoth-
esis. In addition, our results showed that not only ungulates, but
also smaller mammalian herbivores induce rampant indirect effects
in invertebrate food webs, and that such indirect cascading effects
are not restricted to keystone species or individual trophic links,
but hold true across broad and taxonomically diverse trophic groups
within the local food web.
In the following, we first discuss the main interactions between mam-
mals and invertebrate feeding types. Thereafter, we discuss in more detail
the bottom-up cascading effects of mammals across invertebrate trophic
groups. Finally, we discuss the differences in the strength of these cascad-
ing interactions between the short- and tall-grass vegetation, in the light
of the “foraging-intensity” and “tolerance/avoidance” hypotheses.

(a) Pitfall catches (b) Suction samples

B 1,000
B
B
Invertebrate herbivores

300 B
800
B
250 B
600
A
200
400
A
150
200

100

(c) (d)
400 AB
Invertebrate predators

12

10
350 B
8

6 A
300 A
4

250 2

(e) (f)
B 8
B
Invertebrate omnivores

250
A
6
AB
200
4
150

2
100

0
50
No. De. De. De.
(g) Ma. Ma.
Mo.
F I G U R E 5   Invertebrate feeding type
Invertebrate detritivores

90 Mammals excluded
abundances for the different mammal exclusion
80 treatments in short- and tall-grass vegetation.
Left panels: pitfall catches. Right panels: suction
70
samples. (a and b) Invertebrate herbivore
60 abundance. (c and d) Invertebrate predator
abundance. (e and f) Invertebrate omnivore
50 abundance. (g) Invertebrate detritivore
abundance. Almost no invertebrate detritivores
No. De. De. De. (two individuals in total) were caught using
Ma. Ma. suction sampling and are therefore not depicted.
Mo. Symbols, SEs, abbreviations and letters as in
Mammals excluded Figure 4
VANDEGEHUCHTE ET AL. Journal of Animal Ecology |
      1443

T A B L E 1   Results of the general(ized) linear mixed models testing the effects of vegetation type (Vegetation), mammalian herbivore
exclusion treatment (Treatment) and their interaction on the different vegetation characteristics and invertebrate feeding type abundances.
df: degrees of freedom (numerator, denominator). PCO, principal coordinate analysis

Vegetation Treatment Vegetation × Treatment

Variable df F p df F p df F p

Vegetation
N content (%) 1,11 4.16 .066 3,46 0.40 .757 3,46 2.08 .117
Quantity (m2/m) 1,11 9.85 .009 3,47 6.80 .001 3,47 0.19 .902
2 −1 −1
Thickness (m m cm ) 1,11 7.33 .020 3,47 2.12 .111 3,47 2.01 .125
PCO axis 1 scores 1,11 8.84 .013 3,47 3.22 .031 3,47 2.95 .042
Invertebrates pitfall catches
Herbivores 1,11 3.38 .093 3,48 5.02 .004 3,48 0.60 .617
Predators 1,11 4.96 .048 3,48 1.53 .219 3,48 1.08 .365
Omnivores 1,11 4.78 .051 3,48 3.63 .019 3,48 2.35 .084
Detritivores 1,11 0.22 .649 3,48 2.08 .116 3,48 1.40 .254
Invertebrates suction samples
Herbivores 1,11 0.02 .895 3,48 15.61 <.001 3,48 0.79 .504
Predators 1,11 6.58 .026 3,48 4.93 .005 3,48 1.98 .130
Omnivores 1,11 5.54 .038 3,47 1.85 .150 3,47 0.19 .902

Bold p-values are significant at the .05 level.

medium-sized mammals, confirming these invertebrates’ known de-

4.1 | Interactions between mammals and
invertebrate feeding types
In the short-grass vegetation, the abundance of herbivorous inver-
tebrates increased when more forage became available due to the
exclusion of large and smaller-sized mammals, suggesting resource
competition between mammalian and invertebrate herbivores. This
result is in accordance with studies conducted in other ecosystems
showing competition between invertebrate herbivores and large (e.g.
Gómez & González-Megías, 2002; Teichman, Nielsen, & Roland, 2013)
or smaller-sized mammals (e.g. Davidson et al., 2010; Huntzinger et al.,
2008). Higher vegetation quantity and thickness due to mammal ex-
clusions were negatively related to the abundance of ground-dwelling
invertebrate detritivores and predators in the short-grass vegetation,
which could result from a changed microclimate for invertebrates liv-
ing in or close to the litter layer (Lessard, Sackett, Reynolds, Fowler, &
Sanders, 2011; van Klink, van der Plas, van Noordwijk, WallisDeVries,
& Olff, 2015). An alternative explanation is that these invertebrates
move more slowly through the denser vegetation, which may reduce
the probability of being caught in pitfall traps. Vegetation composi-
tion was significantly related to invertebrate herbivore, omnivore
and detritivore abundance, but these links were independent from
mammalian herbivore manipulations. Vegetation composition may
have influenced the vegetation’s structural heterogeneity and avail-
ability of alternative resources for omnivores, such as nectar or pollen
(Langellotto & Denno, 2004; van Rijn, van Houten, & Sabelis, 2002).
However, our experiment did not permit testing this hypothesis.
In the tall-grass vegetation, detritivores positively responded
to increases in vegetation quality caused by excluding large and
pendence on forage quality (Moore et al., 2004; van Klink et al., 2015).
Excluding all or all but small mammals shifted the composition of the
tall-grass vegetation in a way that benefited invertebrate herbivores
but disadvantaged invertebrate detritivores.
4.2 | Bottom-up cascading effects of mammals
across invertebrate trophic levels
In the short-grass vegetation, mammalian herbivore exclusions
caused increases in vegetation quantity that, unlike in the tall-grass
vegetation, initiated strong bottom-up interaction chains across
multiple invertebrate trophic levels. Exclusion-induced increases in
vegetation quantity were linked to increases in herbivorous inver-
tebrate abundance, which in turn benefited predatory and ground-
dwelling omnivorous invertebrates. Although ground-dwelling
invertebrate predators and omnivores were linked to mammals via
several additional pathways, the strong interactions linking mam-
mals, vegetation quantity, invertebrate herbivores and inverte-
brate predators were highly consistent between the ground- and
vegetation-dwelling invertebrate food webs. Regardless of the role
of mammals, bottom-up effects of vegetation properties on inver-
tebrate herbivore and detritivore abundance that cascade up to
invertebrate predators or omnivores have been demonstrated be-
fore (Gruner, 2004; Kagata & Ohgushi, 2006; Rzanny et al., 2013;
Scherber et al., 2010). However, our study is to our awareness
the first to link bottom-up cascades across multispecies inverte-
brate trophic groups to changes in mammalian herbivore pressure.
Moreover, our study demonstrates the importance of smaller
 |
1444      
Journal of Animal Ecology
mammals for the invertebrate food web. The exclusion of smaller
mammals in addition to ungulates caused the strongest changes in
those vegetation properties that were related to invertebrate feed-
ing guilds. Thus, these ubiquitous but rarely considered smaller
mammals not only have a large impact on the vegetation (Davidson
et al., 2010; Howe et al., 2002, 2006; Rebollo et al., 2013), but also
on the invertebrate food web.
In contrast to the short-grass vegetation, we detected only
one bottom-up interaction chain from the vegetation to lower
and higher invertebrate trophic levels in the tall-grass vegetation.
Higher vegetation quality, as observed when large and medium-
sized mammals were excluded, positively related to the abundance
of detritivores, which in turn showed a positive relationship to
ground-dwelling omnivore abundance. However, simultaneous
exclusion-induced shifts in vegetation composition dampened this
cascade through a negative link to detritivore abundance. The lim-
ited and fairly weak cascading effects of mammalian herbivores on
the invertebrate food web in the tall-grass vegetation suggest that
the invertebrates are relatively inert to changes in the abundance
or composition of mammalian grazers in this less intensively grazed
ecosystem. Shifts in mammalian herbivore assemblages that could
arise because of re-immigrating large predators, population man-
agement or local extinctions are therefore not likely to cause major
disruptions of the invertebrate food web at temporal and spatial
scales similar to those of our study. However, a low-productivity
system such as the tall-grass vegetation may be slow to respond,
and permanent changes such as extinctions of mammals may have
effects in the longer term.
Movement of invertebrates into or out of the exclosures may have
contributed to the observed mammal effects on local invertebrate
food webs (cf. Englund, 1997), keeping us from speculating about
the influence of mammals on entire invertebrate populations. Testing
population effects would require excluding mammalian herbivores
from the entire invertebrate habitat, in our case the entire subalpine
grassland. This is practically unfeasible in our study system, but may
be attainable in other systems. While enclosures with realistic densi-
ties of the study organisms, rather than exclosures, solve the problem
of movement into or out of plots (Englund, 1997; Gil et al., 2016), en-
closing mammals and invertebrates together would be impracticable.
Population-level effects of predation can be estimated from observed
prey consumption rates and estimated predator and prey population
sizes (Englund, 1997). However, estimating changes in multitrophic
invertebrate populations from the consumption of plant biomass by
mammalian herbivores would be much less trivial. Moreover, this ap-
proach only focuses on competition-mediated effects of mammals on
invertebrates, ignoring the many other possible mechanisms.
4.3 | “Foraging-intensity” vs. “tolerance/
avoidance” hypothesis
The mammal-induced bottom-up interaction cascades were
stronger in the more productive, more intensively grazed veg-
etation type, in line with the “foraging-intensity” hypothesis. Yet,
VANDEGEHUCHTE ET AL.
Pringle et al. (2007) reported a negative relationship between pro-
ductivity and the strength of cascading effects of ungulate brows-
ing on trees, beetles and insectivorous lizards. Likewise, a recent
global meta-analysis reported that negative indirect effects of large
herbivores on other animals were on average weaker in more pro-
ductive systems, in support of the “tolerance/avoidance” hypothe-
sis (Daskin & Pringle, 2016). Spatial scale and grazing intensity may
help explain this discrepancy with our results. Daskin and Pringle
(2016) related herbivore-induced interaction strength to a global
plant productivity map with a 1-km resolution, and Pringle et al.
(2007) used plots much larger than ours across a large-scale pro-
ductivity gradient. We compared neighbouring areas that differ in
mammalian grazing intensity because the mammals prefer to feed
on the more productive short-grass patches, which are akin to graz-
ing lawns (McNaughton, 1984). A literature study by Chase, Leibold,
Downing, and Shurin (2000) proposed that herbivores should have
weaker effects on plant biomass in ecosystems with higher plant
productivity, because of faster plant regrowth. Yet, when faster
plant regrowth leads to a higher local grazing intensity, mammalian
herbivores can nonetheless have a stronger impact on the biomass
of more productive vegetation (Daskin & Pringle, 2016; Oesterheld
& McNaughton, 2000). Our more productive short-grass vegetation
had higher plant regrowth, but also higher consumption of vegeta-
tion by mammalian herbivores and therefore lower plant biomass
(see Section 2; Risch et al., 2013) than the tall-grass vegetation.
Because of these properties, our system is more likely to conform
to the “foraging-intensity” hypothesis at the scale of our experi-
ment, that is productive vs. less productive vegetation patches
within grasslands. However, we propose that the two hypotheses
are not mutually exclusive. The “foraging-intensity” mechanism
may operate at small scales, particularly when mammals strongly
focus their grazing on productive vegetation patches within the
local landscape, while the “tolerance/avoidance” mechanism may
simultaneously act at larger spatial scales, especially when mamma-
lian herbivore pressure is uniform. Future research could elucidate
the roles of spatial scale and variation in grazing intensity in deter-
mining the balance between “tolerance/avoidance” and “foraging-
intensity” mechanisms.
ACKNOWLEDGEMENTS
We thank Roman Alther, Bieke Boden, Monika Carol Resch,
Charlotte Schaller, Magdalena Steiner, Silvan Stöckli and Peter Wirz
for help with invertebrate sampling and sorting. We also thank vari-
ous employees and volunteers of both the Swiss Federal Institute
for Forest, Snow and Landscape Research and the Swiss National
Park for help with fence construction and maintenance. We thank
the Swiss National Park for permission to work in the park and for
logistic support. We also thank the Associate Editor and two anony-
mous reviewers for constructive comments on previous versions
of the manuscript. This study was funded by the Swiss National
Science Foundation, SNF grant-no 31003A_122009/1 and SNF
grant-no 31003A_140939/1.
VANDEGEHUCHTE ET AL.
AUTHORS’ CONTRIBUTIONS
A.C.R. and M.S. conceived the experiment; A.C.R., F.d.S., M.L.V. and
M.S. collected the data; F.d.S. conducted digital image analysis; M.L.V.
performed statistical analyses; M.L.V. led the writing of the manu-
script. All authors contributed critically to the drafts and gave final
approval for publication.
DATA ACCE SS IBILITY
Data available from the Dryad Digital Repository https://doi.
org/10.5061/dryad.qt2kt (Vandegehuchte, Schütz, de Schaetzen, &
Risch, 2017).
O RC ID
Martijn L. Vandegehuchte http://orcid.org/0000-0003-1283-4654
REFERENCES
Bakker, E. S., Ritchie, M. E., Olff, H., Milchunas, D. G., & Knops, J. M. H.
(2006). Herbivore impact on grassland plant diversity depends on habi-
tat productivity and herbivore size. Ecology Letters, 9, 780–788.
Bardgett, R. D., & Wardle, D. A. (2003). Herbivore-mediated linkages be-
tween aboveground and belowground communities. Ecology, 84,
2258–2268.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate
– A practical and powerful approach to multiple testing. Journal of the
Royal Statistical Society Series B-Methodological, 57, 289–300.
Borer, E. T., Seabloom, E. W., Shurin, J. B., Anderson, K. E., Blanchette, C. A.,
Broitman, B., … Halpern, B. S. (2005). What determines the strength of
a trophic cascade? Ecology, 86, 528–537.
Chase, J. M., Leibold, M. A., Downing, A. L., & Shurin, J. B. (2000). The ef-
fects of productivity, herbivory, and plant species turnover in grassland
food webs. Ecology, 81, 2485–2497.
Cousineau, D. (2005). Confidence intervals in within-subject designs: A sim-
pler solution to Loftus and Masson’s method. Tutorials in Quantitative
Methods for Psychology, 1, 42–45.
Crooks, K. R., & Soule, M. E. (1999). Mesopredator release and avifaunal
extinctions in a fragmented system. Nature, 400, 563–566.
Danell, K., & Hussdanell, K. (1985). Feeding by insects and hares on birches
earlier affected by moose browsing. Oikos, 44, 75–81.
Daskin, J. H., & Pringle, R. M. (2016). Does primary productivity modulate
the indirect effects of large herbivores? A global meta-analysis. Journal
of Animal Ecology, 85, 857–868.
Davidson, A. D., Ponce, E., Lightfoot, D. C., Fredrickson, E. L., Brown, J.
H., Cruzado, J., … Ceballos, G. (2010). Rapid response of a grassland
ecosystem to an experimental manipulation of a keystone rodent and
domestic livestock. Ecology, 91, 3189–3200.
Englund, G. (1997). Importance of spatial scale and prey movements in
predator caging experiments. Ecology, 78, 2316–2325.
Evans, D. M., Villar, N., Littlewood, N. A., Pakeman, R. J., Evans, S. A.,
Dennis, P., … Redpath, S. M. (2015). The cascading impacts of livestock
grazing in upland ecosystems: A 10-year experiment. Ecosphere, 6, 42.
Ford, H., Garbutt, A., Jones, L., & Jones, D. L. (2013). Grazing management
in saltmarsh ecosystems drives invertebrate diversity, abundance
and functional group structure. Insect Conservation and Diversity, 6,
189–200.
Foster, C. N., Barton, P. S., & Lindenmayer, D. B. (2014). Effects of large
native herbivores on other animals. Journal of Applied Ecology, 51,
929–938.
Journal of Animal Ecology |
      1445
Gil, M. A., Jiao, J., & Osenberg, C. W. (2016). Enrichment scale determines
herbivore control of primary producers. Oecologia, 180, 833–840.
Gómez, J. M., & González-Megías, A. (2002). Asymmetrical interactions be-
tween ungulates and phytophagous insects: Being different matters.
Ecology, 83, 203–211.
Grace, J. B. (2006). Structural equation modeling and natural systems.
Cambridge, UK: Cambridge University Press.
Gruner, D. S. (2004). Attenuation of top-down and bottom-up forces in a
complex terrestrial community. Ecology, 85, 3010–3022.
Hairston, N. G., Smith, F. E., & Slobodkin, L. B. (1960). Community structure,
population control, and competition. American Naturalist, 94, 421–425.
Howe, H. F., Brown, J. S., & Zorn-Arnold, B. (2002). A rodent plague on
prairie diversity. Ecology Letters, 5, 30–36.
Howe, H. F., Zorn-Arnold, B., Sullivan, A., & Brown, J. S. (2006). Massive and
distinctive effects of meadow voles on grassland vegetation. Ecology,
87, 3007–3013.
Hunter, M. D., & Price, P. W. (1992). Playing chutes and ladders:
Heterogeneity and the relative roles of bottom-up and top-down
forces in natural communities. Ecology, 73, 724–732.
Huntly, N. (1991). Herbivores and the dynamics of communities and eco-
systems. Annual Review of Ecology and Systematics, 22, 477–503.
Huntzinger, M., Karban, R., & Cushman, H. (2008). Negative effects of
vertebrate herbivores on invertebrates in a coastal dune community.
Ecology, 89, 1972–1980.
Jones, C. G., Lawton, J. H., & Shachak, M. (1997). Positive and negative effects
of organisms as physical ecosystem engineers. Ecology, 78, 1946–1957.
Kagata, H., & Ohgushi, T. (2006). Bottom-up trophic cascades and material
transfer in terrestrial food webs. Ecological Research, 21, 26–34.
Kline, R. B. (2005). Principles and practice of structural equation modeling.
New York, NY: The Guilford Press.
Langellotto, G. A., & Denno, R. F. (2004). Responses of invertebrate natural
enemies to complex-structured habitats: A meta-analytical synthesis.
Oecologia, 139, 1–10.
Lessard, J.-P., Sackett, T. E., Reynolds, W. N., Fowler, D. A., & Sanders, N. J.
(2011). Determinants of the detrital arthropod community structure:
The effects of temperature and resources along an environmental gra-
dient. Oikos, 120, 333–343.
Loranger, H., Weisser, W. W., Ebeling, A., Eggers, T., De Luca, E., Loranger, J.,
… Meyer, S. T. (2014). Invertebrate herbivory increases along an exper-
imental gradient of grassland plant diversity. Oecologia, 174, 183–193.
McNaughton, S. J. (1984). Grazing lawns: Animals in herds, plant form, and
coevolution. American Naturalist, 124, 863–886.
McNaughton, S. J. (1985). Ecology of a grazing ecosystem: The Serengeti.
Ecological Monographs, 55, 259–294.
Milchunas, D. G., & Lauenroth, W. K. (1993). Quantitative effects of grazing
on vegetation and soils over a global range of environments. Ecological
Monographs, 63, 327–366.
Mooney, K. A., Gruner, D. S., Barber, N. A., Van Bael, S. A., Philpott, S. M.,
& Greenberg, R. (2010). Interactions among predators and the cascad-
ing effects of vertebrate insectivores on arthropod communities and
plants. Proceedings of the National Academy of Sciences of the United
States of America, 107, 7335–7340.
Moore, J. C., Berlow, E. L., Coleman, D. C., de Ruiter, P. C., Dong, Q.,
Hastings, A., … Wall, D. H. (2004). Detritus, trophic dynamics and bio-
diversity. Ecology Letters, 7, 584–600.
Morey, R. D. (2008). Confidence intervals from normalized data: A cor-
rection to Cousineau (2005). Tutorials in Quantitative Methods for
Psychology, 4, 61–64.
Nuttle, T., Yerger, E. H., Stoleson, S. H., & Ristau, T. E. (2011). Legacy of
top-down herbivore pressure ricochets back up multiple trophic levels
in forest canopies over 30 years. Ecosphere, 2, 4.
Oesterheld, M., & McNaughton, S. J. (2000). Herbivory in terrestrial eco-
systems. In O. E. Sala, R. B. Jackson, H. A. Mooney, & R. W. Howarth
(Eds.), Methods in ecosystem science (pp. 151–157). New York, NY:
Springer-Verlag.
 |
1446      
Journal of Animal Ecology
Oksanen, L., Fretwell, S. D., Arruda, J., & Niemelä, P. (1981). Exploitation
ecosystems in gradients of primary productivity. American Naturalist,
118, 240–261.
Paine, R. T. (2000). Phycology for the mammalogist: Marine rocky shores
and mammal-dominated communities – How different are the struc-
turing processes? Journal of Mammalogy, 81, 637–648.
Pringle, R. M., Young, T. P., Rubenstein, D. I., & McCauley, D. J. (2007).
Herbivore-initiated interaction cascades and their modulation by pro-
ductivity in an African savanna. Proceedings of the National Academy of
Sciences of the United States of America, 104, 193–197.
Prugh, L. R., & Brashares, J. S. (2012). Partitioning the effects of an ecosys-
tem engineer: Kangaroo rats control community structure via multiple
pathways. Journal of Animal Ecology, 81, 667–678.
Rebollo, S., Milchunas, D. G., Stapp, P., Augustine, D. J., & Derner, J. D.
(2013). Disproportionate effects of non-colonial small herbivores on
structure and diversity of grassland dominated by large herbivores.
Oikos, 122, 1757–1767.
Risch, A. C., Haynes, A. G., Busse, M. D., Filli, F., & Schütz, M. (2013). The
response of soil CO2 fluxes to progressively excluding vertebrate and
invertebrate herbivores depends on ecosystem type. Ecosystems, 16,
1192–1202.
Ritchie, M. E., Tilman, D., & Knops, J. M. H. (1998). Herbivore effects on
plant and nitrogen dynamics in oak savanna. Ecology, 79, 165–177.
Rosenheim, J. A. (1998). Higher-order predators and the regulation
of insect herbivore populations. Annual Review of Entomology, 43,
421–447.
Rzanny, M., Kuu, A., & Voigt, W. (2013). Bottom-up and top-down forces
structuring consumer communities in an experimental grassland. Oikos,
122, 967–976.
Scherber, C., Eisenhauer, N., Weisser, W. W., Schmid, B., Voigt, W., Fischer,
M., … Tscharntke, T. (2010). Bottom-up effects of plant diversity on
multitrophic interactions in a biodiversity experiment. Nature, 468,
553–556.
Schmitz, O. J., Hambäck, P. A., & Beckerman, A. P. (2000). Trophic cascades
in terrestrial systems: A review of the effects of carnivore removals on
plants. American Naturalist, 155, 141–153.
Schütz, M., Risch, A. C., Achermann, G., Thiel-Egenter, C., Page-Dumroese,
D. S., Jurgensen, M. F., & Edwards, P. J. (2006). Phosphorus transloca-
tion by red deer on a subalpine grassland in the Central European Alps.
Ecosystems, 9, 624–633.
Shurin, J. B., Borer, E. T., Seabloom, E. W., Anderson, K., Blanchette, C. A.,
Broitman, B., … Halpern, B. S. (2002). A cross-ecosystem comparison of
the strength of trophic cascades. Ecology Letters, 5, 785–791.
Takagi, S., & Miyashita, T. (2012). Variation in utilization of young leaves by
a swallowtail butterfly across a deer density gradient. Basic and Applied
Ecology, 13, 260–267.
Takagi, S., & Miyashita, T. (2014). Scale and system dependencies of in-
direct effects of large herbivores on phytophagous insects: A meta-
analysis. Population Ecology, 56, 435–445.
VANDEGEHUCHTE ET AL.
Teichman, K. J., Nielsen, S. E., & Roland, J. (2013). Trophic cascades: Linking
ungulates to shrub-dependent birds and butterflies. Journal of Animal
Ecology, 82, 1288–1299.
Terborgh, J., Lopez, L., Nuñez, P., Rao, M., Shahabuddin, G., Orihuela, G.,
… Balbas, L. (2001). Ecological meltdown in predator-free forest frag-
ments. Science, 294, 1923–1926.
van Klink, R., van der Plas, F., van Noordwijk, C. G. E., WallisDeVries, M. F.,
& Olff, H. (2015). Effects of large herbivores on grassland arthropod
diversity. Biological Reviews, 90, 347–366.
van Rijn, P. C. J., van Houten, Y. M., & Sabelis, M. W. (2002). How plants
benefit from providing food to predators even when it is also edible to
herbivores. Ecology, 83, 2664–2679.
Vanbergen, A. J., Hails, R. S., Watt, A. D., & Jones, T. H. (2006). Consequences
for host-parasitoid interactions of grazing-dependent habitat hetero-
geneity. Journal of Animal Ecology, 75, 789–801.
Vandegehuchte, M. L., Raschein, U., Schütz, M., Gwiazdowicz, D. J., & Risch,
A. C. (2015). Indirect short- and long-term effects of aboveground in-
vertebrate and vertebrate herbivores on soil microarthropod commu-
nities. PLoS ONE, 10, 22.
Vandegehuchte, M. L., Schütz, M., de Schaetzen, F., & Risch, A. C. (2017).
Data from: Mammal-induced trophic cascades in invertebrate food
webs are modulated by grazing intensity in subalpine grassland. Dryad
Digital Repository, https://doi.org/10.5061/dryad.qt2kt
Wardle, D. A., Bonner, K. I., & Barker, G. M. (2002). Linkages between plant
litter decomposition, litter quality, and vegetation responses to herbi-
vores. Functional Ecology, 16, 585–595.
Wardle, D. A., Yeates, G. W., Barker, G. M., & Bonner, K. I. (2006). The influ-
ence of plant litter diversity on decomposer abundance and diversity.
Soil Biology & Biochemistry, 38, 1052–1062.
White, T. C. R. (1978). Importance of a relative shortage of food in animal
ecology. Oecologia, 33, 71–86.
Zehm, A., Nobis, M., & Schwabe, A. (2003). Multiparameter analysis of vertical
vegetation structure based on digital image processing. Flora, 198, 142–160.
S U P P O RT I N G I N FO R M AT I O N
Additional Supporting Information may be found online in the
supporting information tab for this article.
How to cite this article: Vandegehuchte ML, Schütz M, de
Schaetzen F, Risch AC. Mammal-induced trophic cascades in
invertebrate food webs are modulated by grazing intensity in
subalpine grassland. J Anim Ecol. 2017;86:1434–1446.
https://doi.org/10.1111/1365-2656.12744