Megaherbivore impacts on ecosystem and Earth system functioning: the current state of the science

Abstract

Megaherbivores (adult body mass > 1000 kg) are suggested to disproportionately shape ecosystem and Earth system functioning. We systematically reviewed the empirical basis for this general thesis and for the more specific hypotheses that 1) megaherbivores have disproportionately larger effects on Earth system functioning than their smaller counterparts, 2) this is true for all extant megaherbivore species and 3) their effects vary along environmental gradients. We furthermore explored possible biases in our understanding of megaherbivore impacts. We found that there are too few studies to quantitatively evaluate the general thesis or any of the hypotheses for all but the African savanna elephant. Following this finding, we performed a qualitative vote counting analysis. Our synthesis of this analysis suggests that megaherbivores can elicit strong impacts on, for example, vegetation structure and biodiversity, and all the elephant species promote seed dispersal. We were, however, unable to evaluate whether these effects are disproportionate to smaller large herbivores. Although environmental conditions can mediate megaherbivore impact, few studies quantified the effect of rainfall or soil fertility on megaherbivore impacts, precluding prediction of megaherbivore effects on the Earth system, particularly under future climates. Moreover, our review highlights major taxonomic, thematic and geographic biases in our understanding of megaherbivore effects. Most of the studies focused on African savanna elephant impacts on vegetation structure and biodiversity, with other megaherbivores and Earth system functions comparatively neglected. Studies were also biased towards semi‐arid and relatively fertile systems, with the arid, high‐rainfall and/or nutrient‐poor parts of the megaherbivores' distribution ranges largely unrepresented. Our findings highlight that the empirical basis of our understanding of the ecological effects of extant megaherbivores is still limited for all species, except the African savanna elephant, and that our current understanding is biased towards certain environmental and geographic areas. We further outline a detailed, urgently needed avenue for future research.

Full text

www.ecography.org
ECOGRAPHY
Ecography
1579
––––––––––––––––––––––––––––––––––––––––
© 2021 e Authors. Ecography published by John Wiley & Sons Ltd on behalf of Nordic Society Oikos
is is an open access article under the terms of the Creative Commons
Attribution License, which permits use, distribution and reproduction in any
medium, provided the original work is properly cited.
Subject Editor: Kate Lyons
Editor-in-Chief: Miguel Araújo
Accepted 2 August 2021
44: 1579–1594, 2021
doi: 10.1111/ecog.05703
44 1579–1594
Megaherbivores (adult body mass > 1000 kg) are suggested to disproportionately
shape ecosystem and Earth system functioning. We systematically reviewed the empiri-
cal basis for this general thesis and for the more specific hypotheses that 1) megaher-
bivores have disproportionately larger effects on Earth system functioning than their
smaller counterparts, 2) this is true for all extant megaherbivore species and 3) their
effects vary along environmental gradients. We furthermore explored possible biases
in our understanding of megaherbivore impacts. We found that there are too few
studies to quantitatively evaluate the general thesis or any of the hypotheses for all
but the African savanna elephant. Following this finding, we performed a qualitative
vote counting analysis. Our synthesis of this analysis suggests that megaherbivores can
elicit strong impacts on, for example, vegetation structure and biodiversity, and all the
elephant species promote seed dispersal. We were, however, unable to evaluate whether
these effects are disproportionate to smaller large herbivores. Although environmental
conditions can mediate megaherbivore impact, few studies quantified the effect of
rainfall or soil fertility on megaherbivore impacts, precluding prediction of megaher-
bivore effects on the Earth system, particularly under future climates. Moreover, our
review highlights major taxonomic, thematic and geographic biases in our understand-
ing of megaherbivore effects. Most of the studies focused on African savanna elephant
impacts on vegetation structure and biodiversity, with other megaherbivores and Earth
system functions comparatively neglected. Studies were also biased towards semi-arid
and relatively fertile systems, with the arid, high-rainfall and/or nutrient-poor parts of
the megaherbivores’ distribution ranges largely unrepresented. Our findings highlight
that the empirical basis of our understanding of the ecological effects of extant mega-
herbivores is still limited for all species, except the African savanna elephant, and that
our current understanding is biased towards certain environmental and geographic
areas. We further outline a detailed, urgently needed avenue for future research.
Keywords: Earth system functioning, ecosystem functioning, herbivore impact,
megaherbivore
Megaherbivore impacts on ecosystem and Earth system
functioning: the current state of the science
Olli Hyvarinen, Mariska Te Beest, Elizabeth le Roux, Graham Kerley, Esther de Groot,
Rana Vinita and Joris P. G. M. Cromsigt
O. Hyvarinen (https://orcid.org/0000-0002-8988-1662) ✉ (olli.hyvarinen@slu.se) and J. P. G. M. Cromsigt, Dept of Wildlife, Fish and Environmental
Studies, Swedish Univ. of Agricultural Sciences, Umeå, Sweden. – JPGMC, M. Te Beest, E. de Groot and R. Vinita, Utrecht Univ., Copernicus Inst. of
Sustainable Development, Utrecht, the Netherlands. – MTB, JPGMC, E. le Roux and G. Kerley, Centre for African Conservation Ecology, Nelson Mandela
Univ., Gqeberha, South Africa. ELR also at: Centre for Biodiversity Dynamics in a Changing World (BIOCHANGE), Section of Ecoinformatics and
Biodiversity, Dept of Biology, Aarhus Univ., Denmark.
Review and Synthesis

1580
Introduction
Large-bodied animals and Earth system functioning
e global climate and biodiversity crises highlight the grow-
ing urgency to better understand the connections and inter-
actions between the different parts of the Earth system. e
Earth system consists of different spheres, such as the atmo-
sphere, geosphere, hydrosphere and biosphere, that are all
interlinked by dynamic and complex processes (Kerényi and
McIntosh 2020, Steffen et al. 2020). A major disruption in
the processes within one sphere can influence processes in
other spheres and, therefore, affect the entire Earth system.
Here, we define ‘Earth system function’ as any process that is
embedded in at least one of these spheres and that supports
the structure and/or stability of the Earth system.
e Earth’s biosphere has shaped the atmosphere and
hydrosphere for at least 2.5 billion years, that is, since the
Great Oxidation Event (Pufahl and Hiatt 2012). Large-
bodied animals are increasingly recognized as playing impor-
tant roles in the functioning of the biosphere and thus the
Earth system (Cromsigt et al. 2018, Schmitz et al. 2018).
eir prehistoric and historic dramatic loss (i.e. defaunation)
has, therefore, been proposed as an underestimated driver of
global change (Estes et al. 2011). A growing body of litera-
ture explores the effects of Pleistocene defaunation on various
Earth system functions (Brault et al. 2013), including the dis-
tribution of biomes (Gill 2014, Doughty et al. 2016c, Dantas
and Pausas 2020), biodiversity (Gill 2014), biogeochemistry
(Doughty et al. 2016c), seed dispersal (Pires et al. 2018), fire
regimes (Gill et al. 2009, Rule et al. 2012), surface energy
fluxes (Doughty et al. 2010, Brault et al. 2013) and pathogen
dispersal (Doughty et al. 2020). Simultaneously, there is an
increasing interest in the ongoing effects of extant large-bod-
ied animals on Earth system functioning (Smith et al. 2016,
Cromsigt et al. 2018, Schmitz et al. 2018). For example,
mammals, as prime dispersers of seeds of certain hardwood
tree species, importantly contribute to the carbon sequestra-
tion potential of tropical forests (Bello et al. 2015). us, a
disruption in seed dispersal (a biosphere process) by defauna-
tion can lead to changes in carbon sequestration (a process that
intersects the atmosphere, biosphere and geosphere). Other
recent examples of how extant large-bodied animals shape
Earth system functioning include reindeer Rangifer tarandus
grazing and trampling reducing shrub cover in the arctic tun-
dra, thereby increasing surface albedo (Te Beest et al. 2016)
and beavers (Castor spp.) changing watershed chemistry and
hydrology (Rosell et al. 2005, Nummi et al. 2018).
Environmental conditions shape the magnitude and
direction of herbivore effects
Environmental conditions are known to mediate the mag-
nitude and direction of the ecological impacts of large her-
bivores on, for example, vegetation structure, soil processes
and fire regimes. For instance, Augustine and McNaughton
(2006) found that the impacts of wild grazers on primary
productivity varied along rainfall and soil fertility gradients.
ey reported that increasing rainfall improved the aboveg-
round productivity on relatively fertile soils while suppressing
it on nutrient-poor soils, leading to different grazing impacts.
Similarly, Waldram et al. (2008) found that white rhino
Ceratotherium simum impact on grassland structure and fire
regimes was more pronounced in the higher rainfall areas of
their study area compared to the lower rainfall areas.
Megaherbivore effects on Earth system functioning
Megaherbivores, as defined by Owen-Smith (1988) are plant-
eating mammals that weigh > 1000 kg as adults. e term
‘megaherbivore’ differs from the increasingly popular term
‘megafauna’, which often refers to animals with adult body
mass > 100 lbs (~45 kg), but the latter is not based on a func-
tional distinction (Moleón et al. 2020). In contrast, their very
large body size distinguishes megaherbivores functionally
from smaller species. First, it renders them near-immune to
non-human predation and top-down population control by
large carnivores. Consequently, megaherbivores are bottom-
up limited by food resources, exacerbating their impact on the
environment (Caughley 1976). Second, owing to their size,
megaherbivores require a large intake of forage, but their low
mass-specific metabolic rate allows them to tolerate low-qual-
ity forage (Müller et al. 2013). As a result, they can consume
more fibrous plant material than smaller species, which leads
to impacts on a wider range of plant species and plant parts
and potentially more homogenous space use. ird, their
size enables megaherbivores to cover greater distances than
smaller species, allowing them to move nutrients and seeds
much further (Owen-Smith 1988, Doughty et al. 2016a).
Because of these functional differences, megaherbivores are
hypothesized to have disproportionately larger effects on eco-
systems than their smaller counterparts (Owen-Smith 1988),
thus eliciting stronger effects on ecosystem and Earth system
functioning than smaller herbivore species, even when occur-
ring at the same biomass density (Fig. 1).
Aims and scope of the study
Here, we systematically review published, peer-reviewed
studies that presented empirical data on contemporary mega-
herbivore effects on ecosystem and Earth system functioning.
While traditional reviews can be useful in summarizing the
state of a scientific discourse, systematic reviews may reveal
and reduce publication and selection bias by deploying a
strict methodology that promotes transparency, objectivity
and repeatability (Haddaway et al. 2015). We are unaware of
any studies that have systematically reviewed the literature on
the ecological and Earth system effects specifically of extant
megaherbivore species. Our main aim was to evaluate the
empirical basis for ecological impacts of megaherbivore spe-
cies and for the thesis that megaherbivores shape the function-
ing of the biosphere (i.e. ecosystems) and the Earth system as
a whole. We also test the more specific, generally assumed
hypotheses that 1) megaherbivores have disproportionately

1581
Figure 1. (A–C) Illustration of potential megaherbivore impacts on various aspects of Earth system functioning: (A) white rhino impact on
vegetation structure, terrestrial biodiversity and fire, (B) hippo impact on vegetation structure, terrestrial biodiversity, biogeochemistry,
hydrology and aquatic biodiversity and (C) African savanna elephant impact on seed dispersal, vegetation structure, terrestrial biodiversity
and fire.

1582
larger effects on Earth system functioning than their smaller
counterparts, that 2) this is true for all megaherbivore species
and that 3) their effects vary along environmental gradients.
Our second aim was to synthesize the current-state-of-the-art
of our understanding of megaherbivore impacts on the Earth
system and to explore possible biases in our understanding.
We evaluated studies that used megaherbivore density or
presence/absence contrasts (hereafter ‘effect contrasts’) and
that report effect sizes (therefore being eligible to be used in
a quantitative meta-analysis) (here classified as type I stud-
ies). We also included more descriptive studies that did not
meet the criteria for a formal quantitative meta-analysis,
such as the reporting of effect sizes and confidence intervals
(here classified as type II studies) (Supporting information).
Following Owen-Smith’s (1988) definition, extant terres-
trial megaherbivore species include African savanna elephant
Loxodonta africana, African forest elephant Loxodonta cyclotis,
Asian elephant Elephas maximus, white rhinoceros, black
rhinoceros Diceros bicornis, greater one-horned rhinoceros
Rhinoceros unicornis, Javan rhinoceros Rhinoceros sondaicus,
common hippopotamus Hippopotamus amphibius as well
as giraffe Giraffa camelopardalis and Sumatran rhinoceros
Dicerorhinus sumatrensis. e latter two species marginally fit
the definition as only some adult individuals exceed the 1000
kg threshold (Table 1).
Material and methods
Study design
We systematically reviewed peer-reviewed empirical studies on
megaherbivore effects on ecosystem and Earth system func-
tioning published between 1945 and 1 July 2020 following the
widely used PRISMA guidelines. ese guidelines describe the
routines and criteria for systematic reviews and meta-analyses
(Moher et al. 2009). We included all extant megaherbivores
in this review (Table 1). We conducted the literature search 1
May 2019 on the Web of Science core collections database and
updated the search 1 July 2020. e search string consisted of
the common and scientific names of all the megaherbivore spe-
cies and terms for effect (Supporting information).
Screening process
First, the search was narrowed by excluding studies not
published in peer-reviewed journals and those not written
in English. All the remaining studies were filtered through
a stepwise screening based on pre-defined relevance and
inclusion criteria (steps 1–3) and quality criteria (step 4)
(Supporting information). Figure 2 gives more details on the
criteria. In step 1, the titles of the publications were evaluated
against criteria set 1, and all the titles deemed irrelevant were
excluded from further analysis. Step 2 exclusions were based
on abstracts evaluated against criteria set 2, and step 3 exclu-
sions were based on the full-text evaluated against criteria set
3. In step 4, we categorized the remaining publications into
type I and type II based on the reported methods and results,
which we evaluated against criteria set 4. Type I publications
consisted of studies that fit the criteria for formal quantitative
meta-analyses (i.e. those that deployed effect contrasts, tested
significance and reported measures of uncertainty) while type
II publications did not have an effect contrast and/or did not
test significance or report measures of uncertainty (i.e. type II
publications were ineligible for quantitative meta-analyses).
For steps 1–3, all studies were evaluated by two assessors inde-
pendently. e lead author, O H, screened through all the
search outputs while co-authors E D and R V each screened
half of the search outputs for steps 1–3. e first half of the
search output consisted of publications on the three elephant
species, while the second half included the rest of the studied
species. In case of a disagreement, all three afore-mentioned
authors discussed the publication in question until an agree-
ment about its inclusion or exclusion was reached. Step 4 was
carried out solely by O H.
Data collection
For each publication that passed the full-text screening
(both type I and type II studies), we recorded the authors,
journal, year of publishing, study location(s), mean annual
temperature, mean annual precipitation, a measure of soil
fertility (cation exchange capacity), each megaherbivore spe-
cies studied and each Earth system function studied. For
type I studies, we further recorded the effect contrast type for
each response variable at the detail reported in the study (i.e.
Table 1. Summary of megaherbivore characteristics. Adult body weight, feeding strategy and gut morphology are extracted from Owen-
Smith (1988), while conservation status and population number are extracted from the IUCN red list (‘The IUCN Red List of Threatened
Species’).
Megaherbivore Adult body
weight (kg)
Feeding
strategy Gut morphology Conservation status
Population
number
African savanna and forest elephants 2500–6000 mixed feeder hind-gut fermenting vulnerable 415 000
Asian elephant 2720–5400 mixed feeder hind-gut fermenting endangered 41 410–52 345
White rhino 1600–2300 grazer hind-gut fermenting near threatened 17 212–18 915
Black rhino 700–1300 browser hind-gut fermenting critically endangered 5630
Greater one-horned rhino 1600–2100 grazer hind-gut fermenting vulnerable 3588
Sumatran rhino 800 browser hind-gut fermenting critically endangered 80
Javan rhino 1300 browser hind-gut fermenting critically endangered 68
Giraffe 800–1200 browser ruminant vulnerable 97 562
Hippopotamus 1365–2600 grazer hind-gut fermenting vulnerable 115 000–130 000

1583
level 1 in Fig. 3, e.g. mortality of Vachellia tortillis < 2 m or
concentration of total phosphorus in soil, etc.), whether the
effect was significant or not based on p-values (significance
cut-off < 0.05) and/or confidence intervals and the direc-
tion of the effect (whether increasing or decreasing) (Fig. 3,
Supporting information). If the effect on the response vari-
able was not significant, it was reported as such (i.e. ‘no sig-
nificant effect’). For type II studies, we further recorded each
measured response variable, but in slightly coarser categories
than for type I studies (i.e. Level 2 in Fig. 3) (e.g. woody cover
or nutrient concentration etc.), and the direction of effect if
applicable. If there was no observed effect, it was reported as
such (i.e. ‘no effect’) (Fig. 3).
Analysis of potential biases
We evaluated both type I and type II publications for taxo-
nomic, thematic, geographic and environmental (tempera-
ture, precipitation and soil fertility) biases. For evaluating
taxonomic and thematic biases, we compared the number of
studies published on the different megaherbivore species and
the different Earth system functions. For this purpose, we
grouped all selected articles into the following seven general
Earth system function categories: vegetation structure, biodi-
versity, biogeochemistry, seed dispersal, fire, hydrology as well
as soil and geomorphology. For evaluating geographic bias,
we first extracted the current and prehistoric distributions for
Figure 2. Prisma flow diagram of the systematic review process including identification, screening eligibility and inclusion of publications.
Reasons for exclusion in each step and the characteristics of type I and type II papers are described in the yellow column on the right.
Figure 3. A schematic overview of the levels of data collection and analysis. For type I publications, we extracted each unique response vari-
able at the finest level (level 1) and further categorized them into a general response variable category (level 2). For type II publications, we
extracted response variables directly at level 2. We finally assigned each response from type I and type II publications into an Earth system
function category at level 3. We performed our qualitative synthesis at level 2, and our analysis of biases at level 3.

1584
each megaherbivore species from the Phylacine database (pre-
historic distributions called ‘present natural’ in the database
of origin (Faurby et al. 2020)). For current distributions, we
only used records that were corroborated by the distribution
estimates reported in Wilson and Reeder (2005) (Supporting
information). We then mapped the publication study sites
and evaluated their geographic locations relative to the cur-
rent and prehistoric distributions of each megaherbivore spe-
cies. To analyze environmental bias, we first extracted the
climate (mean annual precipitation, mean annual tempera-
ture) and elevation data from WorldClim database at 10 min
resolution (Fick and Hijmans 2017). We further derived the
soil fertility data from ISRIC as mean soil cation exchange
capacity at pH 7, 0–5 cm depth at 250 m spatial resolution
(Hengl et al. 2015). We derived the climatic and soil fertility
envelopes for the current distribution of each megaherbivore
species by extracting values for mean annual precipitation,
mean annual temperature and cation exchange capacity from
1000 random points throughout their current distribution
ranges. We then plotted the study sites of the different spe-
cies onto their respective climatic and soil fertility envelopes
to identify areas of the envelopes that had not been studied.
Early in our analysis, we noticed unusually high rainfall val-
ues for some of the random points on the Asian elephant,
African savanna elephant and white and black rhinos current
distribution ranges. Due to the low spatial resolution (96.5
km by 96.5 km at 30° north and 30° south) of the Phylacine
data, high-altitude areas, potentially outside of the species’
current distribution ranges, overestimated the averaged val-
ues per pixel included in our analysis. While elephants have
been recorded at high altitude (Yalden et al. 1986), they are
unlikely to spend a significant amount of time at high alti-
tude (Choudhury 1999), and thus we felt it justified to mask
areas above 2000 m from the current distribution ranges in
order to minimize this distortion due to high-elevation out-
lier rainfall and temperature values. is excluded part of
the Himalayas as well as Ethiopian and Lesotho highlands.
We finally analyzed the temporal trends in the publications
per Earth system function studied and citation bias (citation
counts extracted from Google Scholar on 20 July 2020) by
evaluating the relative contribution of the different study
sites to our understanding of the Earth system effects of each
megaherbivore species separately.
Synthesis
With the initial intention of doing a quantitative meta-analysis
of type I publications, we identified all relevant response vari-
ables within each type I publication at the finest level (level 1
in Fig. 3) and recorded the direction of effect per megaherbi-
vore species, that is, ‘increasing’, ‘decreasing’ or ‘no significant’
effect. For a more inclusive qualitative analysis within which
we could include both type I and type II publications, we fur-
ther classified each response variable in each type I and type II
study into a more general response variable category (level 2
in Fig. 3) and recorded the direction of effect at that level per
megaherbivore species, that is, ‘increasing’, ‘decreasing’ or ‘no’
effect. We then qualitatively synthesized the reported mega-
herbivore effects on the level 2 response categories, across all
type I and type II studies using the so-called ‘vote counting’
method. Using this method, we counted the number of sta-
tistically significant ‘increasing’ and ‘decreasing’ as well as ‘no
(significant)’ effects per response category in order to evalu-
ate the overall effect on that particular category (Vogel et al.
2021, Stewart 2010) (Fig. 3).
Results
Literature identification and screening
By specifying the publication type and language in Web of
Science core collections, we first omitted 3202 symposium
presentations, abstracts, newsletters, books and book chap-
ters, postgraduate theses, reports and other grey literature
as well as 622 peer-reviewed publications that were not
written in English, before running the search. Our speci-
fied search query led to 11 977 peer-reviewed publications
for the period from 1945 to 1 July 2020. We excluded 11
016 publications in the relevance screening of titles (step
1), 415 in the relevance screening of abstracts (step 2) and
306 in the relevance screening of full-text (step 3). After
full text screening, 240 publications remained, which we
subjected to a critical appraisal (step 4) during which we
categorized each remaining study as either type I (144) or
type II (96). In other words, only 3% of the 11 977 studies
from the initial search were deemed relevant (i.e. studied
megaherbivore ecological impacts). Moreover, just 46% of
this 3% deployed appropriate methodology and/or report-
ing (i.e. use of effect contrasts, reporting of effect sizes and
measures of uncertainty) to be eligible for a quantitative
meta-analysis.
In the full-text screening (step 3), the most common rea-
sons for exclusion were that the publication was not specifi-
cally focused on megaherbivore ecological impacts (82% of
306 excluded studies), megaherbivore impacts could not
be distinguished from the impact of other herbivores and/
or environmental variables (11% of 306 excluded studies),
or that the publication was a review (5% of 306 excluded
studies). In the critical appraisal (step 4), the most common
reasons for classifying publications as type II (instead of type
I) were the absence of effect contrast (61% of 96 type II stud-
ies), the absence of required test statistics (23% of 96 type
II studies), insufficient quantitative data (11% of 96 type II
studies) or that the publication was based on modelling with-
out yielding novel data (5% of 96 type II studies) (Supporting
information for a full list of excluded papers in step 3 with
reasons for exclusion).
Characteristics of the peer-reviewed publications
e number of both type I and II studies increased strongly
over the years and appeared in a wide diversity of journals
(Supporting information). e vast majority of studies

1585
(70%) was on African savanna elephants, followed by giraffe
and hippo. e other seven species jointly made up about
10% of studies (Fig. 4). Studies on Asian megaherbivores
were particularly rare, with only 16 on Asian elephant, one
on the greater one-horned rhino and none on the other two
rhino species. Only 10% (14) of the included type I pub-
lications and 7% (7) of type II publications looked at the
effects of two or more megaherbivores in the same system,
of which just one quantified the relative effect sizes for each
species separately. From these initial results, we concluded
that we could not perform rigorous formal quantitative
meta-analyses for any of the species and Earth system func-
tions, except for African savanna elephant effects on vegeta-
tion structure and biodiversity. e sample sizes for the other
species were too small (<5 studies) to meaningfully perform
a similar quantitative analysis for all species. Quantitative
meta-analyses for African savanna elephant effects on veg-
etation structure and biodiversity have already been com-
pleted (Guldemond and Van Aarde 2008, Guldemond et al.
2017). Hence, instead of duplicating these studies on the
savanna elephant, we focused our efforts on qualitative
analyses where we were able to include more studies and all
extant megaherbivore species. In terms of the Earth system
functions, the vast majority of studies looked at vegetation
structure (~65%) and biodiversity (~20%), with relatively
few studies on biogeochemistry and seed dispersal and only
a handful on the other Earth system functions (Fig. 4).
Geographic distribution of studies and potential
environmental biases
e included type I and type II studies originated from 26
different countries (Supporting information) and 105 dif-
ferent study areas (Supporting information). e number of
type I and type II publications per country ranged between
1 and 88, whereas the number of publications per study area
ranged between 1 and 31 (Supporting information). Almost
half of type I studies (40%) came from only five areas in three
countries: Kruger National Park in South Africa (20), Mpala
Research Centre in Kenya (18), Addo Elephant National Park
in South Africa (9), Hluhluwe-iMfolozi Park in South Africa
(8) and Hwange National Park in Zimbabwe (5), while the
same proportion of type II studies (41%) came from ten areas
in seven countries (Supporting information). Major parts of
the extant distribution ranges of all megaherbivore species
lacked any studies on their ecological impacts (Supporting
information).
Collectively, the study sites represented only a fraction of
the climate and soil fertility envelopes of the current distri-
bution ranges of these megaherbivore species (Supporting
information). Studies on African savanna elephant were
strongly biased towards the arid and semi-arid parts of their
distribution range, with 83% of the studies in areas that are
below their distribution range’s median rainfall (Supporting
information). In contrast, Asian elephant studies were biased
Figure 4. Chord diagram showing the proportion of studies published on the effects of the different megaherbivore species on each Earth
system function category. e ‘other’ category includes hydrology as well as soil and geomorphology.

1586
towards the mesic and very wet parts of their range (all studies
coming from areas with mean annual precipitation > 1500
mm (Supporting information). White rhino effects were
studied in only three locations, under semi-arid and mesic
conditions, limited to the parts of their range with relatively
high soil fertility (Supporting information). Studies on black
rhino and giraffe were heavily biased towards relatively cool
(17–18°C and 17–22°C, respectively) (Supporting informa-
tion) and relatively fertile areas (Supporting information). All
black rhino studies were performed under relatively similar
rainfall conditions (mean annual precipitation of 506–760
mm), despite black rhinos occurring over a wide range of
rainfall (Supporting information). Studies on hippo were
concentrated in the drier and relatively more fertile parts of
their range (Supporting information). One outlier study site
was present at the high rainfall end of the hippo’s range (at
2607 mm year−1) but comes from outside of their natural dis-
tribution range (from South America where they were intro-
duced) (Shurin et al. 2020).
Synthesis
We extracted 1259 and 99 responses at level 2 of data col-
lection from type I and type II publications, respectively
(Supporting information for a detailed overview per study),
and further classified them into 26 vegetation structure cat-
egories, 47 biodiversity, 10 biogeochemistry, 4 seed dispersal
and 4 other categories (Fig. 5).
1. Vegetation structure
Most studies dealt with the effects of African savanna ele-
phants on woody species, in general concluding that they
open up the landscape by either increasing woody damage
or mortality or decreasing woody cover, density. Effects of
the African forest and the Asian elephant species on woody
communities were more mixed (votes more spread among
‘increasing’, ‘decreasing’ and/or ‘no’ effects), with much fewer
response categories studied. Similar to the African savanna
elephant, several studies on the browsers, black rhino and
giraffe, found that they generally have negative effects on
woody vegetation, with increased woody damage or mortal-
ity or reduced reproduction, height and abundance. Giraffes
were often reported to have negligible effects on many woody
response categories, therefore suggesting a lack of consensus
on the direction of effect. e majority of studies on the graz-
ers, hippo and white rhino, found them to increase grassland
heterogeneity although the direction of their effects on her-
baceous structure was less clear.
2. Biodiversity
Again, most studies on biodiversity impacts dealt with African
savanna elephant impacts, suggesting that they have variable
effects on most plant groups except succulents, for which
there is a voting bias towards them decreasing species rich-
ness. eir impact on the diversity of other organisms varied
widely, with most votes going to ‘increasing’ or ‘no’ effects.
Notable exceptions are the vertebrate foraging behavior as
well as presence and richness indices, where the vote balance
leaned towards ‘decreasing’ and ‘no’ effects. Studies on the
biodiversity impacts of hippo also varied widely with a rela-
tively equal spread of votes among ‘increasing’, ‘decreasing’
and ‘no’ effects or too low number of votes to draw mean-
ingful conclusions about the direction of their effects. e
number of votes for the biodiversity-related variables studied
in the context of the other megaherbivore species were too
few to make any general conclusions.
3. Biogeochemsitry
Most publications on biogeochemistry studied hippo effects
on nutrient content and concentration of water bodies, pre-
dominantly suggesting nutrient addition, that is, responses
collectively leaning towards ‘no’ and ‘increasing’ effects (yet
with substantial variation among studies and elements).
Collectively, only three studies dealt with biogeochemical
effects of white rhino and African savanna elephant, mostly
showing them to promote soil carbon and lateral nutrient
transport (most responses exhibiting ‘no’ and/or ‘positive’
effects). No studies were done on the effects on biogeochem-
istry by the other species.
4. Seed dispersal
Most studies on megaherbivore effects of seed dispersal
have been done on elephants, particularly on African for-
est elephant and Asian elephant. Overall, these studies show
elephants to increase germination success and decrease ger-
mination time (although a large proportion of studies did
not find (significant) effects). For the three elephant species
combined, there is a vote bias towards positive effects on seed
dispersal.
5. Other
Only a handful of studies dealt with other response cat-
egories. African savanna elephants, white rhino and hippo
reduced fire-related variables (five votes in total), and hippo
reduced soil pore space while increasing geomorphology and
hydrology-related variables (only one vote each).
Discussion
We concluded that the number of peer-reviewed, empirical,
studies is still too small (<5 studies) to run formal quantita-
tive meta-analyses for any of the megaherbivore species and
Earth system functions, except for African savanna elephant
impacts on vegetation structure and biodiversity. However,
our qualitative synthesis suggests that megaherbivores can
have a wide variety of impacts on the different Earth system
functions. Yet, the empirical support for this varies substan-
tially across ecosystem processes, species and systems, sug-
gesting considerable contextual complexity that remains
unexplored. Only a few studies directly quantified the effect
of rainfall or soil fertility on megaherbivore impacts. Given
the paucity of studies, we could not quantify the extent to
which surviving megaherbivore species shape contemporary

1587
ecosystems and Earth system functioning or how this varies
across environmental gradients. ere was also insufficient
evidence to evaluate one of the core hypotheses that mega-
herbivore effects are disproportionate to those of smaller her-
bivores. Moreover, almost half of all type I studies suitable
for future meta-analyses, originated from only three study
areas in South Africa, one in Kenya and one in Zimbabwe
potentially leading to major environmental biases in our cur-
rent understanding.
Low inclusion rates
Most published research on non-elephant megaherbivore
species focused on conservation-oriented topics, such as
Figure 5. (A–C) Summary of the results of the vote counting per megaherbivore species. Green columns indicate increasing effect, yellow
columns no (significant) effect and red columns decreasing effect. e intensity of the colour signifies the number of responses in that
category, but does not necessarily reflect the number of studies in that category.

1588
reproduction, habitat suitability and movement ecology
rather than their ecological impacts. is may be a conse-
quence of the conservation status and generally low popula-
tion sizes of most of these species (Table 1), which promotes
conservation management such as re-introduction and range
expansion. As a result, research on these species focuses on
aspects of their ecology that support these conservation
actions. A second reason for the low inclusion rate of type I
studies was methodological and reporting issues such as the
lack of effect contrasts and/or missing effect sizes and mea-
sures of uncertainty. erefore, we encourage researchers
working on megaherbivore effects to invest in studies that use
a comparative approach (effect contrasts) and to report the
essential statistics for inclusion in future quantitative meta-
analyses. Relocation and range expansion programs provide
fruitful opportunities to study megaherbivore impacts as they
have clear ‘effect contrasts’ i.e. before versus after reintroduc-
tion or range expansion (Landman et al. 2014).
Many studies also failed to (or did not aim to) distinguish
megaherbivore impacts from the impacts of smaller large her-
bivores. For example, exclusion experiment studies often sep-
arated the impact of small and medium-sized herbivores from
that of large herbivores, while making no distinction between
large- and megaherbivores (Dharani et al. 2009, Cassidy et al.
Figure 5. (Continued ).

1589
2013) (but see Ogada et al. 2008 and Charles et al. 2017).
Future exclusion experiments, aimed specifically at discern-
ing the impact of megaherbivores from the impact of other
large herbivores, could benefit from examples, such as the
Kenya Long-Term Exclosure Experiment (KLEE) at Mpala
Research Centre (Young et al. 1997) and the, no longer
standing, exclosure design at Hluhluwe-iMfolozi Park (Van
der Plas et al. 2016). e studies that do discern impacts of
megaherbivores from those of other large herbivores suggest
that these two groups can elicit vastly different effects on veg-
etation structure (Van der Plas et al. 2016) and biodiversity
(Ogada et al. 2008). If, in addition, the intention is to assess
the disproportionality of megaherbivore impact, measures of
biomass density must also be included (Van der Plas et al.
2016). Another approach is to carefully quantify the rela-
tive density of the different taxa and use statistical models
to quantify their relative effects (Smit and Archibald 2019).
Taxonomic bias
We found strong taxonomic bias towards the African savanna
elephant, with a complete absence of qualifying studies on
Asian rhino species (apart from one type II study on greater
one-horned rhino). is bias can be partly explained by
the growing conservation management concerns about the
impacts of confined, growing, African savanna elephant pop-
ulations on vegetation structure and biodiversity, prompting
research in these directions (Guldemond et al. 2017). When
studies are solely motivated by concerns of extremely high or
low megaherbivore population densities, their impacts may
not be studied across their entire density range, but only at
the extremes. is presents another potential bias. Most of
the studies that report decreasing impacts of African savanna
elephant on woody cover (Guldemond and Van Aarde 2008),
for example, come from confined fenced areas with relatively
high elephant population numbers. Although these findings
robustly show that high densities of elephants can decrease
woody cover, they do not necessarily demonstrate that such
impacts are universal across population densities and envi-
ronmental gradients (Guldemond and Van Aarde 2008,
Guldemond et al. 2017 for extensive discussions).
e impact of megaherbivores other than African savanna
elephant has generated less management concern, which may
translate into less research focus on ecological impacts of these
species (although see Heilmann et al. (2006) and Luske et al.
(2009) for the discussion on the impact of black rhino on
euphorbia trees). e lack of studies on ecological impacts by
Asian rhino species may at least partly be explained by their
extremely low population sizes and restricted ranges (and
possibly by the English language restriction on this study).
Management of these species is thus focused on enhancing
their conservation status, stimulating research in directions
such as population ecology and habitat selection, rather than
ecological impacts. Furthermore, ecological impacts of a
species occurring at extremely low population densities are
difficult to study, and results of such studies would likely
suffer from type II error (false negative). In other words,
megaherbivore effects studied at extremely low population
densities do not necessarily reflect the effects they would elicit
at higher densities. is point is particularly relevant for the
Javan and Sumatran rhino. Asian elephant and greater one-
horned rhino do occur in several areas in Asia at densities
that would allow for studies on how they shape Earth sys-
tem functions. We strongly encourage such studies and com-
parative work between Asian and African megaherbivores.
Comparisons between African forest versus Asian elephant
and white rhino versus the greater one-horned rhino seem
to be particularly relevant as they seem functionally similar.
Only 10% of type I studies included in our systematic
review dealt with more than one megaherbivore species in
the same landscape, and only one of them was able to differ-
entiate the relative impacts of the different species (Smit and
Archibald 2019). Many studies for instance recorded herbi-
vore damage on woody plants and associated the damage to
a particular megaherbivore based on the physical attributes
of the damage. Certain megaherbivore species, such as black
rhino and savanna elephant, leave a unique fingerprint on the
damage, making it relatively easy for the researcher to identify
which species caused it. Such studies, however, often did not
quantify the respective megaherbivore visitation rate, popu-
lation density nor employ any other effect contrast (Birkett
2002, Muboko 2015). Exclusion studies, on the other hand,
often combined different megaherbivore species as part of the
same treatment, although without quantifying the relative
species-specific impacts. For example, Charles et al. (2017)
studied the impact of different groups of herbivores on vari-
ous aspects of vegetation structure. While African savanna
elephant and giraffe were both studied, they were included
in the same treatment as ‘megaherbivores’ without teasing
apart their relative impacts. Given this inability to compare
between the impacts of different megaherbivore species and
the taxonomic bias in studies mentioned above, extra caution
should be taken when generalizing ‘megaherbivore impact’
across species. is is particularly relevant given the likely dif-
ferences between the ecological impacts of grazers, such as
white rhino and hippo, and browsers, such as black rhino and
giraffe (Owen-Smith 1988).
Thematic (Earth system function) bias
Our analysis revealed clear thematic biases in the literature.
Changes in vegetation structure and biodiversity were the
most studied Earth system function response categories, par-
ticularly for African savanna elephant, with more emphasis
placed on their impact on woody plants than on the herba-
ceous layer, despite a large proportion of their diet consisting
of grasses (Codron et al. 2011). Strikingly, very few stud-
ies addressed the impact of megaherbivores on soils and soil
microbes, even though their foraging, trampling and other
disturbances are expected to have a large impact on them
(Sitters and Andriuzzi 2019). Our understanding of how
megaherbivores influence biogeochemistry is very limited,
and most of our knowledge comes from studies on hippo’s
role in nutrient transport in riverine systems (Stears et al.

1590
2018, Schoelynck et al. 2019). Although megaherbivores
have been also suggested to play major roles in terrestrial
lateral nutrient transport and ecosystem carbon dynam-
ics (Doughty et al. 2016a), very few have studied this for
extant megaherbivore species (but see le Roux et al. (2018)
and Veldhuis et al. (2018) for white rhino’s role in nutri-
ent transport, as well as Sitters et al. (2020), Wigley et al.
(2020) for African savanna elephant’s role in soil carbon
storage). Megaherbivore effects on seed dispersal have only
been studied in the context of the three elephant species,
particularly for African forest elephant and Asian elephant
(Babweteera et al. 2007, Granados et al. 2017) (although
see Dinerstein (1991) for a description of the potential of
greater one-horned rhino for seed dispersal). Giraffes might
have an important role in pollination and seed dispersal,
although their effects have been largely overlooked in the
literature (but see Fleming et al. 2006). Although megaher-
bivores are frequently said to shape fire regimes (Gill et al.
2009, Rule et al. 2012), we found only three fire-related
studies coming from Hluhluwe-iMfolozi Park (on white
rhino (Waldram et al. 2008)), Kruger National Park (on
African savanna elephant and hippo (Smit and Archibald
2019)) and Mpala Research Centre (on African savanna ele-
phant (Kimuyu et al. 2014)). Furthermore, we found only
one type I study on megaherbivore impacts on ecosystem
hydrology (Dutton et al. 2018), and two type II studies on
soil and geomorphology. No studies were found on surface
energy fluxes, pathogen dispersal or any other Earth system
function (although see Keesing et al. 2013).
Our findings reveal mismatches between literature on the
Earth system effects of Pleistocene megaherbivore extinctions
and the studies on modern effects of extant megaherbivores.
First, the Pleistocene literature links megaherbivore extinc-
tions to increases in fire extent and frequency (Gill et al.
2009, Rule et al. 2012), decreases in surface reflectance
(Doughty et al. 2010, Brault et al. 2013) and pathogen dis-
persal (Doughty et al. 2020) as well as changes in lateral
nutrient diffusion and carbon dynamics (Doughty et al.
2016b). ese connections have not been solidly tested
for the extant megaherbivores, although changes in surface
energy fluxes (Te Beest et al. 2016) and pathogen dispersal
(Berggoetz et al. 2014) have been linked to other large her-
bivores. Surviving megaherbivores, in turn, have been well
linked to changes in vegetation structure, aspects of biodi-
versity and seed dispersal, with much weaker understanding
of their effects on other aspects of earth system function,
such as biogeochemistry, hyrdrology and fire. Second, the
Pleistocene literature often upscales their findings to the
biome or global scale, while studies on modern effects of
extant megaherbivores mostly remain at the local to land-
scape scale. Few studies, however, have modelled the impact
of other large-bodied herbivores on processes such as carbon
emissions (Hempson et al. 2017) and surface energy fluxes
(Te Beest et al. 2016) at a biome or global scale. Bridging
these thematic and scale mismatches will strengthen the
basis for our understanding of megaherbivore effects on
Earth system functioning.
Geographic and environmental biases
We also found substantial geographic bias in the literature
on megaherbivore effects with almost half of type I studies
coming from only five African areas (i.e. Kruger National
Park, Addo Elephant National Park and Hluhluwe-iMfolozi
Park in South Africa and Mpala Research Centre, Kenya and
Hwange National Park, Zimbabwe). ese areas are interna-
tionally well-known for their excellent field research facilities,
exemplifying the importance of governments and the pri-
vate sector continuing to invest in long-term field facilities.
Without the presence of such facilities in these five areas, our
understanding of megaherbivore impacts would undoubtedly
be much poorer. In contrast, a similar proportion of type II
studies came from ten areas, including both African and
Asian countries, demonstrating slightly smaller geographic
bias compared to type I studies.
is enormous overall geographic bias, however, poten-
tially leads to further environmental biases in our under-
standing of megaherbivore impacts. Our findings reveal, for
all megaherbivores species, that current study areas only rep-
resent small parts of the climate and soil fertility envelopes
of their current distribution ranges. Both type I and type II
studies are generally biased towards semi-arid and relatively
fertile systems, with a near absence of studies under arid, high
rainfall and nutrient-poor conditions. Furthermore, less than
a handful of studies directly quantified the effect of rainfall or
soil fertility on megaherbivore impacts (Waldram et al. 2008,
Goheen et al. 2013, Smit and Archibald 2019). erefore,
we do not know how megaherbivores shape ecosystems and
Earth system processes for particularly the drier and wetter
parts of their ranges, or how environmental drivers influence
the direction and strength of their effects. e few studies
that we have on megaherbivores and those on other large
herbivores, however, suggest that environmental drivers do
mediate herbivore impacts (Waldram et al. 2008). ese lim-
itations hinder our efforts to predict how future climates may
influence the Earth system effects of megaherbivores.
Emerging trends in megaherbivore impact research
Although much of the research on modern megaherbivore
impacts focus independently on either vegetation structure
or biodiversity, some have recently studied their interac-
tive effects on both vegetation structure and biodiversity
(Ogada et al. 2008), therefore, incorporating a focus on eco-
logical cascades. Although very few studies, in general, were
on megaherbivore impacts on ecosystem biogeochemistry,
two recent publications reported on the impact of white
rhino on lateral nutrient transport, demonstrating their
ability to move nutrients against fear-driven gradients (le
Roux et al. 2018, Veldhuis et al. 2018). ere is an increas-
ing interest in the role of hippos on allochthonous nutri-
ent transport, and their further effects on aquatic primary
productivity and biodiversity. Schoelynck et al. (2019), for
instance, demonstrated that hippos can significantly con-
tribute to the global cycling of silicon by feeding on riverine

1591
grasslands and defecating in water bodies, with potential
cascading impacts on the silicon-limited estuarine diatoms.
Recent studies have also linked African savanna elephants to
changes in above- and below-ground carbon, paving the way
to an exciting research avenue on megaherbivore impacts on
global carbon cycling and carbon sequestration. Interestingly
Sitters et al. (2020) and Wigley et al. (2020) found contrast-
ing impacts of African savanna elephant on soil carbon in the
same system, the former showing an increase and the latter a
decrease in total organic carbon. In addition to investigating
total organic carbon, future research should look at megaher-
bivore effects on the different soil carbon fractions (Lehmann
and Kleber 2015) to better understand how they influence
soil carbon stabilization processes and therefore carbon resi-
dence times. An exciting, although nearly untouched, area of
research is the impact of megaherbivores on terrestrial and
aquatic microorganisms. Our knowledge is limited to but a
few studies that investigated for instance the impact of African
savanna elephant dung on mycorrhizal colonization of plants
(Paugy et al. 2004) and the impact of hippo dung on biofilm
productivity and respiration (Subalusky et al. 2018).
Study limitations
We acknowledge that our focus on English language peer-
reviewed journals may have limited our sample size. With
this systematic review, however, we specifically aimed to assess
the state of the empirical peer-reviewed literature and explic-
itly excluded non-peer-reviewed studies. To identify a pos-
sible language bias in our results, we did a posthoc assessment
using Web of Science core collections database (15 October
2020), which revealed that the total number of studies on
megaherbivores published in other major language peer-
reviewed journals, that is, French, Spanish and Portuguese,
was 342. In this assessment, we ran the search for the sci-
entific names and the common names of the megaherbivore
species in each language, excluding the term of effect, there-
fore making our estimate conservative. Using the 3% inclu-
sion rate based on relevance and the 46% inclusion rate based
on quality (the rates that we found for our original screen-
ing of articles in English), we estimated that we may have
missed five type I and five type II studies published in French,
Spanish or Portuguese language peer-reviewed journals that
would qualify for inclusion in our analysis. is gives us con-
fidence that our systematic review captured a representative
sample of the current global scientific discourse (that is based
on peer -reviewed empirical literature) on the extant mega-
herbivore effects on ecosystem and Earth system functioning.
e small number of qualifying studies that reported the
impacts of a megaherbivore species on a particular response
category (apart from African savanna elephant impact on some
aspects of vegetation structure and biodiversity) prevented us
from running a full quantitative meta-analysis. Instead, we
synthesized the literature through a qualitative ‘meta-analysis’
using the vote-counting method. Although vote-counting is
used widely in the field of applied ecology, it has been criticized
for ignoring sample size and effect magnitude. Researchers
who use vote-counting often synthesize unweighted averages
of effect sizes, when only study estimates but not variances are
available (Stewart 2010). is can lead to bias, because it ignores
the different volumes of information coming from studies of
different size and quality (Stewart 2010). In contrast, we used
vote-counting to qualitatively synthesize the impacts of the
different megaherbivore species on a given response category
(level 2). Instead of synthesizing unweighted effect sizes, we
simply looked at the direction of the effect (increasing, decreas-
ing, no effect). While this approach still ignores the size and
quality of the study, it avoids the pitfall of using unweighted
effect sizes. Despite these shortcomings, vote-counting allowed
us to synthesize the overall impacts of the different megaherbi-
vore species on a given response category, through identifying
areas of agreement and dispute.
Concluding remarks
Our systematic review revealed that the empirical support
for the thesis that extant megaherbivores (>1000 kg) shape
ecosystem and Earth system functioning relies on very few,
localized, studies and suffers from major taxonomic, thematic,
geographic and environmental biases. is prevented us from
running a strictly quantitative meta-analysis for any other spe-
cies than the African savanna elephant. erefore, we could
not evaluate our follow-up hypotheses and thus it remains
largely unclear whether 1) megaherbivores have dispropor-
tionately larger effects on Earth system functioning compared
to their smaller counterparts, and how effects may vary among
2) species and 3) environmental gradients. Despite these
shortcomings, our qualitative ‘meta-analysis’ revealed widely
varying, context and species-dependent impacts of megaher-
bivores on the different response categories. Furthermore,
interesting research avenues are gradually opening on the cas-
cading effects of megaherbivores connecting different Earth
system functions, and a few studies already report on mega-
herbivore effects on micro-organisms, nutrient transport and
carbon cycling. Future research should, however, considerably
increase the number of empirical studies on the ecological
and Earth system effects of the different non-African savanna
elephant megaherbivore species such as African rhino spp and
hippo, and test the net effects of possible interactions among
sympatric megaherbivore species. Furthermore, we must stra-
tegically expand the geographic distribution of studies across
environmental gradients. Finally, we call for more, creative,
studies that aim at differentiating megaherbivore effects from
those of smaller large herbivores.
Acknowledgement – We thank Dr Kees Rookmaker for his advice
regarding Asian rhino literature.
Funding – is research project was funded by the Swedish
Research Council for Sustainable Development, Formas, under the
project acronym Megaclim (diary no. 2017-01000). EL was funded
by the Claude Leon Foundation and through a Royal Society
Newton International Fellowship.
Conflicts of interest – All the authors declare that they have no
competing interests.

1592
Author contributions
Olli Hyvarinen: Conceptualization (lead); Data cura-
tion (lead); Formal analysis (lead); Investigation (lead);
Methodology (lead); Resources (lead); Software (lead);
Validation (lead); Visualization (lead); Writing – original
draft (lead); Writing – review and editing (lead). Mariska te
Beest: Conceptualization (equal); Data curation (support-
ing); Formal analysis (supporting); Investigation (equal);
Methodology (equal); Resources (equal); Supervision
(equal); Visualization (supporting); Writing – original draft
(equal); Writing – review and editing (equal). Liza Roux:
Conceptualization (supporting); Data curation (supporting);
Formal analysis (equal); Investigation (equal); Methodology
(equal); Resources (equal); Software (equal); Supervision
(equal); Validation (equal); Visualization (equal); Writing –
original draft (equal); Writing – review and editing (equal).
Graham I. H. Kerley: Conceptualization (supporting); Data
curation (supporting); Investigation (equal); Methodology
(equal); Supervision (equal); Validation (equal); Writing –
original draft (equal); Writing – review and editing (equal).
Esther de Groot: Data curation (equal); Methodology (sup-
porting). Rana Vinita: Data curation (equal); Methodology
(supporting). Joris P. G. M. Cromsigt: Conceptualization
(lead); Data curation (supporting); Formal analysis (sup-
porting); Funding acquisition (lead); Investigation (equal);
Methodology (equal); Project administration (lead);
Resources (equal); Software (equal); Supervision (lead);
Validation (equal); Visualization (supporting); Writing –
original draft (equal); Writing – review and editing (equal).
Transparent Peer Review
e peer review history for this article is available at <https://
publons.com/publon/10.1111/ecog.05703>.
Data availability statement
Data are available via the Dryad Digital Repository: <https://
doi.org/10.5061/dryad.2z34tmpn4> (Hyvarinen et al. 2021).
References
Augustine, D. J. and McNaughton, S. J. 2006. Interactive effects of
ungulate herbivores, soil fertility and variable rainfall on ecosys-
tem processes in a semi-arid savanna. – Ecosystems 9: 1242–1256.
Babweteera, F. et al. 2007. Balanites wilsoniana: regeneration with
and without elephants. – Biol. Conserv. 134: 40–47.
Bello, C. et al. 2015. Defaunation affects carbon storage in tropical
forests. – Sci. Adv. 1: e1501105.
Berggoetz, M. et al. 2014. Tick-borne pathogens in the blood of
wild and domestic ungulates in South Africa: interplay of game
and livestock. – Ticks Tick-Borne Dis. 5: 166–175.
Birkett, A. 2002. e impact of giraffe, rhino and elephant on the
habitat of a black rhino sanctuary in Kenya. – Afr. J. Ecol. 40:
276–282.
Brault, M.-O. et al. 2013. Assessing the impact of late Pleistocene
megafaunal extinctions on global vegetation and climate. –
Clim. Past 9: 1761–1771.
Cassidy, L. et al. 2013. Effects of restriction of wild herbivore move-
ment on woody and herbaceous vegetation in the Okavango
Delta Botswana. – Afr. J. Ecol. 51: 513–527.
Caughley, G. 1976. Plant–herbivore systems. – In: May, R. M.
(ed.), eoretical ecology. WB Saunders Co., Philadelphia, PA.
Charles, G. K. et al. 2017. Herbivore effects on productivity vary
by guild: cattle increase mean productivity while wildlife reduce
variability. – Ecol. Appl. 27: 143–155.
Choudhury, A. 1999. Status and conservation of the Asian elephant
Elephas maximus in north-eastern India. – Mammal Rev. 29:
141–174.
Codron, J. et al. 2011. Landscape-scale feeding patterns of African
elephant inferred from carbon isotope analysis of feces. – Oec-
ologia 165: 89–99.
Cromsigt, J. P. et al. 2018. Trophic rewilding as a climate change
mitigation strategy? – Phil. Trans. R. Soc. B 373: 20170440.
Dantas, V. and Pausas, J. 2020. e legacy of Southern American
extinct megafauna on plants and biomes. – Authorea Preprints
10.22541/au.160147523.36636469/v2.
Dharani, N. et al. 2009. Browsing impact of large herbivores on
Acacia xanthophloea Benth in Lake Nakuru National Park,
Kenya. – Afr. J. Ecol. 47: 184–191.
Dinerstein, E. 1991. Seed dispersal by greater one-horned rhinoc-
eros Rhinoceros unicornis and the flora of Rhinoceros latrines.
– Mammalia 55: 355–362.
Doughty, C. E. et al. 2010. Biophysical feedbacks between the
Pleistocene megafauna extinction and climate: the first human-
induced global warming? – Geophys. Res. Lett. 37: 15.
Doughty, C. E. et al. 2016a. Global nutrient transport in a world
of giants. – Proc. Natl Acad. Sci. USA 113: 868–873.
Doughty, C. E. et al. 2016b. Megafauna extinction, tree species
range reduction and carbon storage in Amazonian forests. –
Ecography 39: 194–203.
Doughty, C. E. et al. 2016c. e impact of the megafauna extinc-
tions on savanna woody cover in South America. – Ecography
39: 213–222.
Doughty, C. E. et al. 2020. Megafauna decline have reduced path-
ogen dispersal which may have increased emergent infectious
diseases. – Ecography 43: 1107–1117.
Dutton, C. L. et al. 2018. e influence of a semi-arid sub-catch-
ment on suspended sediments in the Mara River, Kenya. –
PLoS One 13: e0192828.
Estes, J. A. et al. 2011. Trophic downgrading of planet Earth. – Sci-
ence 333: 301–306.
Faurby, S. et al. 2020. MegaPast2Future/PHYLACINE_1.2: PHY-
LACINE Ver. 1.2.1. – Zenodo, doi:10.5281/zenodo.3690867.
Fick, S. E. and Hijmans, R. J. 2017. WorldClim 2: new 1-km
spatial resolution climate surfaces for global land areas. – Int.
J. Climatol. 37: 4302–4315.
Fleming, P. A. et al. 2006. Are giraffes pollinators or flower preda-
tors of Acacia nigrescens in Kruger National Park, South Africa?
– J. Trop. Ecol. 22: 247–253.
Gill, J. L. 2014. Ecological impacts of the late Quaternary mega-
herbivore extinctions. – New Phytol. 201: 1163–1169.
Gill, J. L. et al. 2009. Pleistocene megafaunal collapse, novel plant
communities and enhanced fire regimes in North America. –
Science 326: 1100–1103.
Goheen, J. R. et al. 2013. Piecewise disassembly of a large-herbivore
community across a rainfall gradient: the UHURU experiment.
– PLoS One 8: e55192.
Granados, A. et al. 2017. Defaunation and habitat disturbance
interact synergistically to alter seedling recruitment. – Ecol.
Appl. 27: 2092–2101.

1593
Guldemond, R. and Van Aarde, R. 2008. A meta-analysis of the
impact of African elephants on savanna vegetation. – J. Wildl.
Manage. 72: 892–899.
Guldemond, R. A. et al. 2017. A systematic review of elephant
impact across Africa. – PLoS One 12: e0178935.
Haddaway, N. R. et al. 2015. Making literature reviews more reli-
able through application of lessons from systematic reviews:
making literature reviews more reliable. – Conserv. Biol. 29:
1596–1605.
Heilmann, L. C. et al. 2006. Will tree euphorbias (Euphorbia
tetragona and Euphorbia triangularis) survive under the impact
of black rhinoceros Bicornis diceros minor browsing in the Great
Fish River Reserve, South Africa? – Afr. J. Ecol. 44: 87–94.
Hempson, G. P. et al. 2017. e consequences of replacing wildlife
with livestock in Africa. – Sci. Rep. 7: 1–10.
Hengl, T. et al. 2015. Mapping soil properties of Africa at 250 m
resolution: random forests significantly improve current predic-
tions. – PLoS One 10: e0125814.
Hyvarinen, O. et al. 2021. Data from: Megaherbivore impacts on
ecosystem and Earth system functioning: the current state of the
science. – Dryad Digital Repository, <https://doi.org/10.5061/
dryad.2z34tmpn4>.
IUCN 2021. e IUCN Red List of reatened Species. Ver.
2021-1. – <https://www.iucnredlist.org>.
Keesing, F. et al. 2013. Effects of wildlife and cattle on tick abun-
dance in central Kenya. – Ecol. Appl. 23: 1410–1418.
Kerényi, A. and McIntosh, R. W. 2020. Structure and operation of
the Global Society (Anthroposphere). – In: Sustainable develop-
ment in changing complex earth systems. Springer, pp. 203–226.
Kimuyu, D. M. et al. 2014. Native and domestic browsers and
grazers reduce fuels, fire temperatures and acacia ant mortality
in an African savanna. – Ecol. Appl. 24: 741–749.
Landman, M. et al. 2014. Long-term monitoring reveals differing
impacts of elephants on elements of a canopy shrub commu-
nity. – Ecol. Appl. 24: 2002–2012.
le Roux, E. et al. 2018. Megaherbivores modify trophic cascades
triggered by fear of predation in an african savanna ecosystem.
– Curr. Biol. 28: 2493.e3–2499.e3.
Lehmann, J. and Kleber, M. 2015. e contentious nature of soil
organic matter. – Nature 528: 60–68.
Luske, B. L. et al. 2009. Impact of the black rhinoceros Diceros bicornis
minor on a local population of Euphorbia bothae in the Great Fish
River Reserve, South Africa. – Afr. J. Ecol. 47: 509–517.
Moher, D. et al. 2009. Preferred reporting items for systematic
reviews and meta-analyses: the PRISMA statement. – PLoS
Med. 6: e1000097.
Moleón, M. et al. 2020. Rethinking megafauna. – Proc. R. Soc. B
287: 20192643.
Muboko, N. 2015. e role of man, hand-raised black rhinos and
elephants on woody vegetation, Matusadona National Park,
Zimbabwe. – Pachyderm 56: 72–81.
Müller, D. W. et al. 2013. Assessing the Jarman–Bell principle: scal-
ing of intake, digestibility, retention time and gut fill with body
mass in mammalian herbivores. – Comp. Biochem. Physiol. A
Mol. Integr. Physiol. 164: 129–140.
Nummi, P. et al. 2018. Beavers affect carbon biogeochemistry: both
short-term and long-term processes are involved. – Mammal
Rev. 48: 298–311.
Ogada, D. L. et al. 2008. Impacts of large herbivorous mammals
on bird diversity and abundance in an African savanna. – Oec-
ologia 156: 387.
Owen-Smith, R. N. 1988. Megaherbivores: the influence of very
large body size on ecology. – Cambridge Univ. Press.
Paugy, M. et al. 2004. Elephants as dispersal agents of mycorrhizal
spores in Burkina Faso. – Afr. J. Ecol. 42: 225–227.
Pires, M. M. et al. 2018. Pleistocene megafaunal extinctions and
the functional loss of long-distance seed-dispersal services. –
Ecography 41: 153–163.
Pufahl, P. K. and Hiatt, E. E. 2012. Oxygenation of the Earth’s
atmosphere–ocean system: a review of physical and chemical
sedimentologic responses. – Mar. Petrol. Geol. 32: 1–20.
Rosell, F. et al. 2005. Ecological impact of beavers Castor fiber and
Castor canadensis and their ability to modify ecosystems. –
Mammal Rev. 35: 248–276.
Rule, S. et al. 2012. e aftermath of megafaunal extinction: eco-
system transformation in Pleistocene Australia. – Science 335:
1483–1486.
Schmitz, O. J. et al. 2018. Animals and the zoogeochemistry of the
carbon cycle. – Science 362: 6419.
Schoelynck, J. et al. 2019. Hippos Hippopotamus amphibius: the
animal silicon pump. – Sci. Adv. 5: eaav0395.
Shurin, J. B. et al. 2020. Ecosystem effects of the world’s largest
invasive animal. – Ecology 101: e02991.
Sitters, J. and Andriuzzi, W. S. 2019. Impacts of browsing and
grazing ungulates on soil biota and nutrient dynamics. – In:
e ecology of browsing and grazing II. Springer, pp. 215–236.
Sitters, J. et al. 2020. Negative effects of cattle on soil carbon and
nutrient pools reversed by megaherbivores. – Nat. Sustain. 3:
360–366.
Smit, I. P. and Archibald, S. 2019. Herbivore culling influences
spatio-temporal patterns of fire in a semiarid savanna. – J. Appl.
Ecol. 56: 711–721.
Smith, F. A. et al. 2016. Megafauna in the Earth system. – Ecog-
raphy 39: 99–108.
Stears, K. et al. 2018. Effects of the hippopotamus on the chemis-
try and ecology of a changing watershed. – Proc. Natl Acad.
Sci. USA 115: E5028–E5037.
Steffen, W. et al. 2020. e emergence and evolution of Earth sys-
tem science. – Nat. Rev. Earth Environ. 1: 54–63.
Stewart, G. 2010. Meta-analysis in applied ecology. – Biol. Lett. 6:
78–81.
Subalusky, A. L. et al. 2018. Organic matter and nutrient inputs
from large wildlife influence ecosystem function in the Mara
River, Africa. – Ecology 99: 2558–2574.
Te Beest, M. et al. 2016. Reindeer grazing increases summer albedo
by reducing shrub abundance in Arctic tundra. – Environ. Res.
Lett. 11: 125013.
Van der Plas, F. et al. 2016. Different-sized grazers have distinctive
effects on plant functional composition of an African savannah.
– J. Ecol. 104: 864–875.
Veldhuis, M. P. et al. 2018. Spatial redistribution of nutrients by
large herbivores and dung beetles in a savanna ecosystem. – J.
Ecol. 106: 422–433.
Vogel, S. M. et al. 2021. Joining forces toward proactive elephant
and rhinoceros conservation. – Conserv. Biol. 2021, doi:
10.1111/cobi.13726.
Waldram, M. S. et al. 2008. Ecological engineering by a mega-
grazer: white rhino impacts on a South African savanna. – Eco-
systems 11: 101–112.
Wigley, B. J. et al. 2020. Grasses continue to trump trees at soil
carbon sequestration following herbivore exclusion in a semi-
arid African savanna. – Ecology 101: e03008.

1594
Wilson, D. E. and Reeder, D. M. 2005. Mammal species of the
world: a taxonomic and geographic reference. – JHU Press.
Yalden, D. W. et al. 1986. Catalogue of the mammals of Ethiopia:
6. Perissodactyla, proboscidea, hyracoidea, lagomorpha, tubuli-
dentata, sirenia and cetacea: pubblicazioni del centro di studio
per la faunistica ed ecologia tropicali del cnr: cclxxxv. – Monit.
Zool. Ital. Suppl. 21: 31–103.
Young, T. P. et al. 1997. KLEE: a long-term multi-species herbivore
exclusion experiment in Laikipia, Kenya. – Afr. J. Range Forage
Sci. 14: 94–102.