Content uploaded by Sandra Diaz
Author content
All content in this area was uploaded by Sandra Diaz on Apr 03, 2015
Content may be subject to copyright.
ARTICLE doi:10.1038/nature14324
Global effects of land use on local
terrestrial biodiversity
Tim Newbold
1,2
*, Lawrence N. Hudson
3
*, Samantha L. L. Hill
1,3
, Sara Contu
3
, Igor Lysenko
4
, Rebecca A. Senior
1
{, Luca Bo
¨rger
5
,
Dominic J. Bennett
4
{, Argyrios Choimes
3,4
, Ben Collen
6
, Julie Day
4
{, Adriana De Palma
3,4
, Sandra Dı
´az
7
,
Susy Echeverria-London
˜o
3
, Melanie J. Edgar
3
, Anat Feldman
8
, Morgan Garon
4
, Michelle L. K. Harrison
4
, Tamera Alhusseini
4
,
Daniel J. Ingram
4
{, Yuval Itescu
8
, Jens Kattge
9,10
, Victoria Kemp
4
, Lucinda Kirkpatrick
4
{, Michael Kleyer
11
,
David Laginha Pinto Correia
3
, Callum D. Martin
4
, Shai Meiri
8
, Maria Novosolov
8
, Yuan Pan
4
, Helen R. P. Phillips
3,4
,
Drew W. Purves
2
, Alexandra Robinson
4
, Jake Simpson
4
, Sean L. Tuck
12
, Evan Weiher
13
, Hannah J. White
4
{, Robert M. Ewers
4
,
Georgina M. Mace
6
,Jo
¨rn P. W. Scharlemann
1,14
& Andy Purvis
3,4
Human activities, especially conversion and degradation of habitats, are causing global biodiversity declines. How local
ecological assemblages are responding is less clear—a concern given their importance for many ecosystem functions and
services. We analysed a terrestrial assemblagedatabase of unprecedented geographic and taxonomic coverage to quantify
local biodiversity responses to land use and related changes. Here we show that in the worst-affected habitats, these
pressures reduce within-sample species richness by an average of76.5%, total abundance by 39.5% andrarefaction-based
richness by 40.3%. We estimate that, globally, these pressures have already slightly reduced average within-sample
richness (by 13.6%), total abundance (10.7%) and rarefaction-based richness (8.1%), with changes showing marked
spatial variation. Rapid further losses are predicted under a business-as-usual land-use scenario; within-sample
richness is projected to fall by a further 3.4% globally by 2100, with losses concentrated in biodiverse but economically
poor countries. Strong mitigation can deliver muchmore positive biodiversity changes (up to a 1.9% average increase) that
are less strongly related to countries’ socioeconomic status.
Biodiversity faces growing pressures from human actions, includ-
ing habitat conversion anddegradation, habitat fragmentation, climate
change, harvesting and pollution
1
. As a result, global assessments show
that species’ extinction risk is increasing on average while population
sizes are declining
1,2
. Such assessments have usually focused on data-
rich vertebrates, so might not reflect broader biodiversity
3
. Furthermore,
most have concentrated on the global status of species, whereas the
long-term security of many ecosystem functions and services – especially
in changing environments – is likely to depend upon local biodiver-
sity
4–6
. Average trendsin local diversity remainunclear: analyses of tem-
poral changes in assemblages have suggested no systematic change in
species richness
7,8
, but the available times-series data might under-
represent transitions between land-use types
9
, and population time series
suggest vertebrate populations have declined sharply in recent decades
3
.
Spatialcomparisons provide an alternative source of evidence on how
human pressures affect biodiversity, assuming that differences in pres-
sures have caused observed biodiversity differences between otherwise
matched sites
10–12
. The prevalence of published spatial comparisons makes
it possible to go beyond particular taxa or regions
11,12
to develop global,
taxonomically representative models. Furthermore, the willingness of
many researchers to share their raw data makes it possible to consider
multiple aspects of biodiversity, rather than the single, simple metrics
of most existing models
10
, which cannot capture all key aspects of
diversity
13
.
We present the most geographically and taxonomically represent-
ative models to date of how several aspects of the composition and
diversity of terrestrial assemblages respond to multiple human pres-
sures. The explanatory variables in our models most directly measure
land use and infrastructure, but might correlate
14,15
with two other im-
portant pressures, harvesting and invasive species, for which compar-
able high-resolution spatial data are unavailable globally. We exclude
climatechange effects because they are not captured well by spatialcom-
parisons. We use our models to infer past net changes in assemblages
since the year 1500, project future changes over this century under dif-
ferentsocioeconomicscenarios of land use,and relate projected national
changes in local biodiversity to socioeconomic variables and natural
biodiversity.
Our models of local within-sample species richness (hereafter ‘rich-
ness’), rarefaction-basedspecies richness (hereafter ‘rarefied richness’),
total abundance, compositional turnover and average organism size are
*These authors contributed equally to this work.
1
United Nations Environment Programme World Conservation Monitoring Centre, 219 Huntingdon Road, Cambridge CB3 0DL, UK.
2
Computational Science Laboratory, Microsoft Research Cambridge, 21
Station Road, Cambridge CB1 2FB, UK.
3
Department of Life Sciences, Natural History Museum, Cromwell Road, London SW7 5BD, UK.
4
Department of Life Sciences, Imperial College London, Silwood Park,
London SL5 7PY, UK.
5
Department of Biosciences, College of Science, Swansea University, Singleton Park, Swansea SA2 8PP, UK.
6
Department of Genetics, Evolution and Environment, Centre for
Biodiversity and Environment Research, University College London, Gower Street, London WC1E 6BT, UK.
7
Instituto Multidisciplinario de Biologı
´a Vegetal (CONICET-UNC) and FCEFyN, Universidad
Nacional de Co
´rdoba, Casilla de Correo 495, 5000 Co
´rdoba, Argentina.
8
Deptartment of Zoology, Faculty of Life Sciences, Tel-Aviv University, 6997801 Tel Aviv, Israel.
9
Max Planck Institute for
Biogeochemistry, Hans Kno
¨ll Straße 10, 07743 Jena, Germany.
10
German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Deutscher Platz 5e, 04103 Leipzig, Germany.
11
Landscape
Ecology Group, Institute of Biology and Environmental Sciences, University of Oldenburg, D-26111 Oldenburg, Germany.
12
Department of Plant Sciences, University of Oxford, Oxford OX1 3RB, UK.
13
Biology Department, University of Wisconsin–Eau Claire, Eau Claire, Wisconsin 54701, USA.
14
School of Life Sciences, University of Sussex, Brighton BN1 9QG, UK. {Present addresses: Department of
Animal and Plant Sciences, University of Sheffield, Alfred Denny Building, Western Bank, Sheffield S10 2TN, UK (R.A.S.); Department of Earth Science and Engineering, Imperial College London, London
SW7 2AZ, UK and Institute of Zoology, Zoological Society of London, London NW1 4RY, UK (D.J.B.); College of Life and Environmental Sciences, Hatherly Laboratories, University of Exeter, Prince of Wales
Road, Exeter EX4 4PS, UK (J.D.); School of Life Sciences, University of Sussex, Brighton BN1 9QG, UK (D.J.I.); School of Biological and Ecological Sciences,University of Stirling, Stirling FK9 4LA, UK (L.K.);
School of Biological Sciences, Queen’s University Belfast, 97 Lisburn Road, Belfast BT9 7BL, UK (H.J.W.).
G2015 Macmillan Publishers Limited. All rights reserved
2APRIL2015|VOL520|NATURE|45
based on among-site comparisons of ecological assemblage composi-
tion collated from the literature as part of the PREDICTS project
16
.
The data set consisted of 1,130,251 records of abundance and 320,924
of occurrence or species richness at 11,525 sites (2–360 sites per study,
median 15; Fig. 1a). These data, from 284 publications (see Methods),
represent 26,953 species (1.4% of the number formallydescribed
17
) and
13 of the 14 terrestrial biomes (Extended Data Fig. 1). Each site was
scored for six putative pressures: land use
11
and use intensity
18
, land-
use history
19
, human population density
20
, proximity to roads
21
and
accessibility from the nearest large town. Random effects in our mod-
els accounted for study-level differences in response variables and sam-
pling methods, and for the within-study spatial arrangement of sites.
Effects of pressure on site-level diversity
Local richness, rarefied richness and total abundance were most strongly
influenced by land use and land-use intensity: they were substantially
lower in mostother land-usetypes than in primaryvegetation, especially
in intensivelyused areas (Fig. 1; see SupplementaryInformation for sta-
tistics and coefficientestimates). These results extend those of previous,
geographically or taxonomically restricted, meta-analyses (for example,
refs 11, 22). Other variables were weaker as main effects, but showed
stronger effects in interaction (Extended Data Fig. 2) and were often
significant overall (see Supplementary Information). Richness and total
abundance tended to be slightly lower at the highest human population
densities, and richness was lower nearer to roads andin more accessible
sites (Fig. 1). Differences in richness were not driven solely by differ-
ences in abundance. Rarefied richness
23
(see Methods for details) showed
weaker but mostly similar patterns, although the effects of variables other
than land use and land-use intensity were not significant (Extended
Data Fig. 3a, b). Under the worst combinations of pressures, our models
estimated richness, rarefied richness and total abundance to be 76.5%,
40.3% and 39.5% lower, respectively, than in minimally affected sites.
Effects of pressures on vertebrate, invertebrate and plant richness were
statistically indistinguishable (P.0.05; results not shown). The mod-
elled coefficients were robust to efforts to correct for any publication
bias (Extended Data Fig. 4). As with all studies based on data from the
literature, unpublished data are almostunrepresented. Coefficients were
also robust under cross-validation (Extended Data Fig. 3c, d), and the
model residuals showed little spatial autocorrelation (Extended Data
Fig. 5).
The importance of secondary vegetation for conservation is a hotly
debated topic
11,24,25
, and an important one, given that this land use will
soon become the most widespread type
26
. We find that the answer
depends strongly on the secondary vegetation’s maturity: early-stage
communities tend to be less diverse than those in primary vegetation
and are compositionallydistinct, but thesedifferences aremuch reduced
in mature secondary vegetation (Figs 1 and 2; we caution though that
not all data sources clearly distinguished mature secondary from prim-
ary vegetation). This successional rise in diversity accords with a recent
meta-analysis of plant communities over time
7
.
Net changes in diversity provide an incomplete view of the effects of
human activities on biodiversity because they ignore the replacement
of original species by newcomers
8
. We therefore analysed how land use
affects similarity in species composition between sites. Communities
under the same land use were, as expected, the most similar (Fig. 2a).
Across land uses, communities in primary vegetation were most like
those in secondary vegetation, while plantation forest, pasture and crop-
land communities formed a different, human-dominated cluster (Fig. 2b).
Anthropogenic pressures can affectecosystem functions and services
more strongly than changes in species diversity would imply, if species’
responses depend on their traits
27
. Large size is often linked to species’
declines
28,29
and is important for some ecosystem processes
30
. We com-
bined abundance data with species’ average sizes to calculate site-level
community-weighted mean plant height and animal mass. As in local
studies
29
, mean plant height was lower in human-dominated land uses
than in primary and secondary vegetation, and tended to decline with
increasing human population density (Fig. 1d). Most field studies focused
on particular plant taxa, so this difference does not simply reflect tree
removal. Average animal mass did not change consistently with land
use or human population density, but increased with proximity to roads
(Fig. 1d).
Models like ours that substitute space for time ignore time lags in
biotic changes, which can be important
31
. We also a ssume that land uses
a
−80
−60
−40
−20
0
20
40
Richness difference (%)
Primary
MSV
ISV
YSV
Plantation
Cropland
Pasture
Urban
HPD
PR
ACC
L
M
H
L
M
H
L
M
H
bLand-use intensity:
Minimal
Light
Intense
Light and intense
−100
−50
0
50
100
Abundance difference (%)
Primary
MSV
ISV
YSV
Plantation
Cropland
Pasture
Urban
HPD
PR
ACC
L
M
H
c
−100
−50
0
50
100
150
CWM size difference (%)
Primary
Secondary
Plantation
Cropland
Pasture
HPD
PR
L
M
HL
M
H
Plant height
Animal mass
d
Figure 1
|
Locations of sites and responses of four metrics of local
diversity to human pressures. a, Sites used in the models. b–d, Responses
44
of richness (b), total abundance (c) and community-weighted mean
(CWM) organism size—plant height (crosses) and animal mass (triangles)—
(d) to anthropogenic variables. Error bars show 95% confidence intervals.
Primary, primary vegetation; YSV, young secondary vegetation;
ISV, intermediate secondary vegetation; MSV, mature secondary vegetation;
plantation, plantation forest. Land-use intensity is categorized as minimal
(circle), light (triangle), intense (diamond), or combined light and intense
(square). HPD, human population density
45
; PR, proximity to roads
46
(as
2log(distance to nearest road)); and ACC, accessibility to humans
47
(as
2log(travel time to nearest major city)), are shown as fitted effects from a
model with no interactions between continuous effects and land use, at the
lowest (L), median (M) and highest (H) values in the data set. Sample sizes
are given in full in the Methods.
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved
46|NATURE|VOL520|2APRIL2015
are situated randomly within studies relative to sites’ intrinsicsuitability
for biodiversity. Adding global data on other important pressures as
they become available, and also incorporating climate change, will give
a more complete picture of human effects on local biodiversity.
Global effects on local diversity to date
By applying our model for within-sample species richness—the most
widely used and understood biodiversity measure—to maps of current
pressure variables
10
, we estimated the global pattern of net local changes
to date in plot-level richness (Fig. 3; we did not estimate total richness
within the 0.5u30.5ugrid cells). Human-dominated areas are inferred
to have lost much more local diversity than have regions where more
natural vegetation remains. The worst-affected cells showed a 31% reduc-
tion in average localrichness—probably enough to alter ecosystem func-
tioning substantially
4
. Local richness increased in 1.7% of cells (by #4.8%).
Total abundance and, less strongly, rarefied richness showed broadly
similar patterns (Extended Data Fig. 6).
We applied our models to global spatial estimates of how land use
and human population changed from 1500–2005 (ref. 26) (see Methods)
to infer the global history of local biodiversity change. Here we focus on
within-sample species richness because of its wide use and easy inter-
pretation. Our inferences incorporate uncertainty in model parameter
estimates, but not in the trajectories of the pressures themselves(which
have not been assessed
32
) nor effects of changes in roads and accessi-
bility, for which temporal estimates could not be obtained.
Richness is estimated to have declined most rapidly in the 19th and
20th centuries (Fig. 4), with other metrics showing similar responses
(Extended Data Fig. 6). By 2005, we estimate that land use and related
pressures had reduced local richness by an average of 13.6% (95% con-
fidence interval (CI): 9.1–17.8%) and total abundance by 10.7% (95%
CI: 3.8% gainto 23.7% reduction) compared with what they would have
been in the absence of human effects. Approximately 60% of the
decline in richness was independent of effects on abundance; average
rarefied richness has fallen by 8.1% (95% CI: 3.5–12.9%). Although
these confidence limits omit uncertainty in the projections of land use
and otherpressures, thereis less uncertainty in estimates of current pres-
sure levels than in changes over time
33
.
Our inferences contrast withtwo recent analyses of community time
series
7,8
, which suggested no overall trend in local diversity, and with the
Living Planet Index
3
, which, based on vertebrate population time se-
ries, reports a much more rapid decline in abundance than we infer.
Although time series potentially provide a more direct view of tem-
poral trends thanour space-for-time approach, the available data might
under-represent transitions between land-use types
9
. However, our ap-
proach may underestimate additions of species through climate change
and species invasion (although accessibility and proximity to roads may
partly capture the latter
14,15
).
Primary
MSV
ISV
YSV
Plantation
Cropland
Pasture
Urban
Primary
MSV
ISV
YSV
Plantation
Cropland
Pasture
Urban 12 0 1114212
49 711 16 716 41 2
32 12 8 14 7 26 16 4
59 10 11 9 54 771
44 13 23 40 9 14 16 1
39 10 35 23 11 811 1
29 24 10 13 10 12 7 0
189 29 39 44 59 32 49 12
a
−0.49
−0.15
−0.02
0.03
0.09
0.12
0.18
0.2
0.23
0.3
0.47
Plantation
Pasture
Cropland
Primary
YSV
MSV
ISV
−0.3
−0.2
−0.1
0.0
0.1
0.2
0.3
b
5%
0%
−32%
Figure 3
|
Net change in local richness caused
by land use and related pressures by 2000.
Projections used an IMAGE reference scenario
10
.
The baseline landscape was assumed to be entirely
uninhabited, unused primary vegetation. Shown
using a Lambert Cylindrical Equal-Area projection
at 0.5u30.5uresolution.
Figure 2
|
Similarity in assemblage composition as a function of land use.
a, Average dissimilarity of species composition (1 2Sørenson Index) between
pairs of sites within and among land uses (shown relative to the similarity
between pairs of primary-vegetation sites); blue and red colours indicate,
respectively, more or less similar composition; numbers indicate numbers of
studies within which comparisons could be made. b, Clustering of land-use
types based on average compositional dissimilarity; urban sites were excluded
owing to the small sample size.
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
2APRIL2015|VOL520|NATURE|47
Global and national projections to 2095
Global changes in local diversity from 2005 to 2095 were projected using
estimated land use and human population from the four Intergovern-
mental Panel on Climate Change Representative Concentration Path-
way (RCP) scenarios
26
, which correspond to different intensities of
global climate change (Table 1). Although these estimates have limita-
tions
32
, they are the most consistent available, are widely used
34
, and
are consistent with the historical estimates
26
. However, they—like all
other global land-use projections—include no estimate of uncertainty;
therefore, each of our projections must be viewed as the predicted bio-
diversity outcome under one particular set of land-use assumptions.
Projected net changes in average local diversity to 2095 vary widely
among scenarios (Fig. 4 and Extended Data Fig. 6). The scenario with
the least climate change (IMAGE 2.6) yields the second-worst out-
come for biodiversity, because it assumes rapid conversion of primary
vegetation, especiallyin the tropics, to crops and biofuels
26
(Table 1 and
Extended Data Fig. 7). These projections do not imply that low-emission
scenarios must entail large losses of biodiversity, but insteadreflect that
scenario’s mitigation strategy. Indeed, in MiniCAM 4.5 (where mit-
igation is through carbon markets, crop improvements and diet shifts,
Table 1) average richness is projected to increase (though other divers-
ity metrics respond more weakly, Extended Data Fig. 6).The worst bio-
diversity outcomes arise from the scenario with most climate change
(MESSAGE 8.5) in which rapid human population growth drives wide-
spread agricultural expansion (Table 1 and Extended Data Fig. 7). This
scenario, which has been characterized as ‘business-as-usual’
35
, most
closely matches recent trends in emissions
36
and gives the worst out-
comes even though our projections omit direct climate effects on local
assemblages.
The global projections hide wide regional and national variation
(Fig. 5 and Extended Data Fig. 8).Projections for 2095 under business-
as-usual (MESSAGE 8.5) are strongly inequitable, presenting serious
challenges for bothsustainable development and global conservation of
biodiversity (Fig. 5a). Under this scenario, European and North Amer-
ican countries, typically with a high Human Development Index, low
native biodiversity and widespread historical land conversion, are mostly
projected to gain in local richness by 2095. More naturally biodiverse
but less economically developed Southeast Asian and especially sub-
Saharan African countries, with more natural and semi-natural hab-
itat, will suffer the greatest losses (Fig. 5a and Extended Data Fig. 8f).
Such globally inequitable outcomes might be avoidable. The bestsce-
nario for biodiversity(MiniCAM 4.5; Fig. 4) yielded country-level out-
comes that are relatively independent of Human Development Index,
native species richness (Fig. 5b) and past changes (Extended Data Fig. 8e).
For local richness, outcomes under MiniCAM4.5 were better than
MESSAGE 8.5 for 93% of countries worldwide (Fig. 5c).
Under AIM 6.0, most Afrotropical countries are projected to gain in
local richness but heavy losses are inferred for the Indo-Malay region
(Extended Data Fig. 8). Projectionsunder IMAGE 2.6 are spatially sim-
ilar to those under MESSAGE 8.5. The land-use change caused by the
biofuels-based strategy in IMAGE 2.6 is projected to have a major neg-
ative effect overall on terrestrial biodiversity (Extended Data Fig. 8).
Conclusions
Many assessments of the state of biodiversity have focused on global
metrics such as rates of speciesextinction
37
, but resilient deliveryof eco-
system functions and services is more likely to depend on local divers-
ity
4–6
. Our models suggest land-use changes and associated pressures
strongly reduce local terrestrial biodiversity, and we estimate global
average reductions to date of 13.6% in within-sample species richness,
10.7% in total abundance and 8.1% in rarefaction-based species rich-
ness (Figs 3 and 4). Climate change, which we could not include in our
−20
−15
−10
−5
0
5
Net richness change (%)
1500 1600 1700 1800 1900 2000 2100
Year
HYDE
IMAGE 2.6
MINICAM 4.5
AIM 6.0
MESSAGE 8.5
Figure 4
|
Projected net change in local richness from 1500 to 2095. Future
projections were based on the four RCP scenarios (Table 1). Historical
(shading) and future (error bars) uncertainty is shown as 95% confidence
intervals, rescaled to zero in 2005. The baseline for projections is a world
entirely composed of uninhabited, unused primary vegetation; thus, the value
at 1500 is not constrained to be zero because by then non-primary land
uses were present (and in some regions widespread). The global average
projection for MESSAGE8.5 does not join the historical reconstructionbecause
that scenario’s human population projections start in 2010 and because
human population and plantation forest extent have not been harmonized
among scenarios.
Table 1
|
Key features of the four RCP scenarios
Scenario Land use (see also Extended Data Fig. 7) Climate and energy Human population
IMAGE 2.6 Agriculture moves from developed to developing
countries. Large increase in area of biofuel plantations.
Urban extent assumed constant.
Increased energy efficiency. Increased use of carbon
capture and storage, nuclear, renewable energy
and biofuels. Approximately 1 uC temperature increase
by 2100 compared to pre-industrial.
10.1 billion by 2100 (UN
Medium variant, 2010)
MiniCAM 4.5 Carbon pricing leads to preservation of primary forest
and expansion of secondary forest. Crop yield
increases, improved agricultural efficiency and dietary
shifts lead to decreases in cropland and pasture
areas. Small increase in area of biofuel plantations.
Urban extent assumed constant.
Decline in overall energy use. Decreased use of fossil
fuels and increase in nuclear and renewable energy,
and in carbon capture and storage. Moderate increase
in use of biofuels, but limited by availability of biomass.
Approximately 1.75 uC temperature increase by 2100.
Peaks above 9 billion around
2065, then declines to 8.7
billion by 2095.
AIM 6.0 Urban area increases owing to human population
growth. Cropland area increases to meet food
demand. Pasture area declines strongly.
Approximately 2.5 uC temperature increase by 2100. 9.1 billion by 2100
(UN Medium variant, 2004)
MESSAGE 8.5 Increasing crop yields and intensification account
for much of the increased production required, but
area of cropland and, to a lesser extent, pasture
increases rapidly. Small increase in area of biofuel
plantations. Urban area increases owing to
increased population.
Small improvements in efficiency leading to high
demand for energy. Conventional oil and gas become
scarce, leading to shift in favour of unconventional
and carbon-intensive fossil fuels. Moderate increase in
use of biofuels. Approximately 4 uC increase in
temperature by 2100.
12 billion by 2100.
Land-use and human population assumptions are detailed in ref. 26, energy assumptions in refs 40–42, and climate implications in ref. 43.
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved
48|NATURE|VOL520|2APRIL2015
framework, is likely to exacerbate losses, especially under business-as-
usual
38
, although direct effects of climate change will increase local
diversity in some regions
8
.
It is important to remember that the habitat conversion and assoc-
iated changes that reduced local biodiversity had largely positive con-
sequences for people; agricultural intensification underpinned many
countries’ development. However, benefits have not been shared equally
among or within countries
39
. Losses of local species richness exceeding
20% are likely to substantially impair the contribution of biodiversity to
ecosystem function and services, and thus to human well-being
4
.We
estimate that reductions in averageplot-level species richness currently
exceed this level for 28.4% of grid cells, increasing to 41.5% of cells by
2095 under business-as-usual (note that we do not estimate or project
total richness across the cell). Importantly, our projections suggest that
such widespread large losses are not inevitable. With concerted action
and the right societal choices, global sustainability of local biodiversity
may be an achievable goal.
Online Content Methods, along with any additional Extended Data display items
and SourceData, are available in theonline version of the paper;references unique
to these sections appear only in the online paper.
Received 9 July 2014; accepted 12 February 2015.
1. Tittensor, D. P. et al. A mid-term analysis of progress toward international
biodiversity targets. Science 346, 241–244 (2014).
2. Pimm, S. L. et al. The biodiversity of species and their rates of extinction,
distribution, and protection. Science 344, 1246752 (2014).
3. Collen, B. et al. Monitoring change in vertebrate abundance: the Living Planet
Index. Conserv. Biol. 23, 317–327 (2009).
4. Hooper, D. U. et al. A global synthesis reveals biodiversity loss as a major driver of
ecosystem change. Nature 486, 105–108 (2012).
5. Isbell,F. et al. High plant diversity is neededto maintain ecosystem services.Nature
477, 199–202 (2011).
6. Cardinale, B. J. et al. Biodiversity loss and its impact on humanity. Nature 486,
59–67 (2012).
7. Vellend, M. et al. Global meta-analysis reveals no net change in local-scale plant
biodiversity over time. Proc. Natl Acad. Sci. USA 110, 19456–19459 (2013).
8. Dornelas, M. et al. Assemblage time series reveal biodiversity change but not
systematic loss. Science 344, 296–299 (2014).
9. Cardinale, B. Overlooked local biodiversity loss. Science 344, 1098 (2014).
10. Alkemade, R. et al. GLOBIO3: a framework to investigate options for reducing
global terrestrial biodiversity loss. Ecosystems 12, 374–390 (2009).
11. Gibson, L. et al. Primary forests are irreplaceable for sustaining tropical
biodiversity. Nature 478, 378–381 (2011).
12. Mendenhall, C. D., Karp, D. S., Meyer, C. F. J., Hadly, E. A. & Daily, G. C. Predicting
biodiversity change and averting collapse in agricultural landscapes. Nature 509,
213–217 (2014).
13. Pereira, H. M. et al. Essential biodiversity variables. Science 339, 277–278 (2013).
14. Weber, E. & Li, B. Plant invasions in China: what is to be expected in the wake of
economic development? Bioscience 58, 437–444 (2008).
15. Clements, G. R. et al. Where and how are roads endangering mammals in
Southeast Asia’s forests? PLoS ONE 9, e115376 (2014).
16. Hudson, L. N. et al. The PREDICTS database: a global database of how local
terrestrial biodiversity responds to human impacts. Ecol. Evol. 4, 4701–4735
(2014).
17. Chapman, A. D. Numbers of Living Species in Australia and the World. (Australian
Biological Resources Study, 2009).
18. Phalan, B., Onial, M., Balmford, A. & Green, R. E. Reconciling food production and
biodiversityconservation: land sharing and landsparing compared. Science 333,
1289–1291 (2011).
19. Balmford, A. Extinction filters and current resilience: the significance of past
selection pressures for conservation biology. Trends Ecol. Evol. 11, 193–196
(1996).
20. Newbold, T. et al. A global model of the responseof tropical and sub-tropicalforest
biodiversity to anthropogenic pressures. Proc. R. Soc. B 281, 20141371 (2014).
21. Benı
´tez-Lo
´pez, A., Alkemade, R. & Verweij, P. A. The impacts of roads and other
infrastructure on mammal and bird populations: a meta-analysis. Biol. Conserv.
143, 1307–1316 (2010).
22. Murphy, G. E. P. & Romanuk, T. N. A meta-analysis of declines in local species
richness from human disturbances. Ecol. Evol. 4, 91–103 (2014).
23. Magurran, A. E. Measuring Biological Diversity. (Wiley-Blackwell, 2004).
24. Barlow, J. et al. Quantifying the biodiversity value of tropical primary, secondary,
and plantation forests. Proc. Natl Acad. Sci. USA 104, 18555–18560 (2007).
25. Dent, D. H. & Wright, S. J. The future of tropical species in secondary forests: A
quantitative review. Biol. Conserv. 142, 2833–2843 (2009).
26. Hurtt, G. C. et al. Harmonization of land-use scenarios for the period 1500–2100:
600 years of global gridded annual land-use transitions, wood harvest, and
resulting secondary lands. Clim. Change 109, 117–161 (2011).
27. Dı
´az, S. et al. Functional traits, the phylogeny of function, and ecosystem service
vulnerability. Ecol. Evol. 3, 2958–2975 (2013).
−30
−20
−10
0
10
20
MESSAGE 8.5
Richness change 2005−2095 (%)
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Human Development Index
a
−30
−20
−10
0
10
20
MINICAM 4.5
Richness change 2005−2095 (%)
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Human Development Index
Palearctic
Afrotropic
Neotropic
Australasia
Indo-Malay
Nearctic
Oceania
b
−15
−10
−5
0
5
10
15
20
MESSAGE 8.5 vs. MINICAM 4.5
Richness change (%) MINICAM 4.5
−30 −20 −10 0 10
Richness chan
g
e (%) MESSAGE 8.5
c
Figure 5
|
Biodiversity projections at the country level. a,b, Country-level
projections of average net local richness change between 2005 and 2095 under
the worst (a, MESSAGE 8.5) and best (b, MiniCAM 4.5) RCP scenarios for
biodiversity, shown in relation to countries’ Human Development Index.
Colours indicate biogeographic realms; colour intensity reflects natural
vertebrate species richness (more intense colour representshigher richness);
point diameter is proportional to (log) country area. c, Correlation between
projected richness changes under the MiniCAM 4.5 and MESSAGE 8.5
scenarios, with dashed line showing equality; colours as in aand b; colour
intensity is proportional to the Human Development Index (more intense
colour represents higher index).
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
2 APRIL 2015 | VOL 520 | NATURE | 49
28. Cardillo, M. et al. Multiple causes of high extinction risk in large mammal species.
Science 309, 1239–1241 (2005).
29. Mayfield, M. M. et al. Differences in forestplant functional trait distributions across
land-use and productivity gradients. Am. J. Bot. 100, 1356–1368 (2013).
30. Se
´guin, A., Harvey, E
´., Archambault, P., Nozais, C. & Gravel, D. Body size as a
predictorof species loss effect on ecosystemfunctioning. Sci. Rep. 4, 4616(2014).
31. Wearn, O. R., Reuman, D. C. & Ewers, R. M. Extinction debt and windows of
conservationopportunity in the Brazilian Amazon. Science 337, 228–232 (2012).
32. Harfoot, M. et al. Integrated assessment models for ecologists: the presentand the
future. Glob. Ecol. Biogeogr. 23, 124–143 (2014).
33. Ellis, E. C. Anthropogenic transformation of the terrestrialbiosphere. Phil. Trans. R.
Soc. A 369, 1010–1035 (2011).
34. Mora, C. et al. The projected timing of climate departure from recent variability.
Nature 502, 183–187 (2013).
35. Burrows, M. T. et al. Geographical limits to species-range shifts are suggested by
climate velocity. Nature 507, 492–495 (2014).
36. Oldfield, F. & Steffen, W. Anthropogenic climate change and the nature of Earth
System science. Anthr. Rev. 1, 70–75 (2014).
37. Pereira, H. M. et al. Scenarios for global biodiversity in the 21st century. Science
330, 1496–1501 (2010).
38. Warren, R. et al. Quantifying the benefit of early climate change mitigation in
avoiding biodiversity loss. Nature Clim. Chang. 3, 678–682 (2013).
39. Millennium Ecosystem Assessment. Ecosystems and Human Well-being:
Biodiversity Synthesis. (World Resources Institute, 2005).
40. van Vuuren, D. P. et al. RCP2.6: exploring the possibility to keep global mean
temperature increase below 2uC. Clim. Change 109, 95–116 (2011).
41. Thomson, A. M. et al. RCP4.5: a pathway for stabilization of radiative forcing by
2100. Clim. Change 109, 77–94 (2011).
42. Riahi, K. et al. RCP 8.5—A scenario of comparatively high greenhouse gas
emissions. Clim. Change 109, 33–57 (2011).
43. Rogelj, J., Meinshausen, M. & Knutti, R. Global warming under old and new
scenarios using IPCC climate sensitivity range estimates. Nature Clim. Chang. 2,
248–253 (2012).
44. Bates, D., Maechler, M., Bolker, B. & Walker, S. lme4: Linear mixed-effects
models using Eigen and S4. http://cran.r-project.org/web/packages/lme4/
(2013).
45. Center for International Earth Science Information Network (CIESIN) Columbia
University, International Food Policy Research Institute(IFPRI), The World Bank &
Centro Internacional de Agricultura Tropical (CIAT). Global rural-urban mapping
project, version 1 (GRUMPv1): population density grid. (NASA Socioeconomic
Data and Applications Center (SEDAC), 2011). http://dx.doi.org/10.7927/
H4R20Z93 (Accessed 11 July 2012).
46. Center for International Earth Science Information Network (CIESIN) Columbia
University & Information Technology Outreach Services (ITOS) University of
Georgia. Global roads open access data set, version 1 (gROADSv1). (NASA
Socioeconomic Data and Applications Center (SEDAC), 2013). http://dx.doi.org/
10.7927/H4VD6WCT (Accessed 18 December 2013).
47. Nelson, A. Estimated travel time to the nearest city of 50,000 or more people in year
2000. http://bioval.jrc.ec.europa.eu/products/gam/index.htm (2008).
(Accessed 14 July 2014).
Supplementary Information is available in the online version of the paper.
Acknowledgements We thank all the many researchers who have made their data
available to us; S. Butchart and Birdlife International for sharing bird body-size data;
F. Gilbert for hoverflybody-size data; the IMAGE,HYDE, MESSAGE and MiniCAM teams,
especially R. Alkemade, M. Bakkenes and A. Thomson for sharing additional data from
their integrated assessment models; D. Tittensor for statistical advice; C. Sleep and
S. Patlola at the Natural History Museum in London for IT support with the database;
members of the GARD initiative (http://www.gardinitiative.org/index.html) for help
with estimating the reptile species richnessmap; K. Jones, J. Tylianakis,M. Crawleyand
E. J. Milner-Gulland for discussion,N. Burgess for commentson a draft of the paper. We
also thank C. D. Thomas and two anonymous reviewers for very helpful comments on
the manuscript.This study is part of the PREDICTS(Projecting Responsesof Ecological
Diversity in Changing Terrestrial Systems) project, which is supported by the UK
Natural Environment Research Council (NERC, grant number: NE/J011193/1), the
Biotechnology and Biological Sciences Research Council (grant number: BB/
F017324/1) a Hans Rausing PhD scholarship. The study was also supported by the
TRY initiative on plant traits, whose database is maintained at Max-Planck-Institute for
Biogeochemistry, Jena, Germany, and which is supported by DIVERSITAS, IGBP, the
Global Land Project, NERC, the French Foundation for Biodiversity Research, and GIS
‘Climat, Environnement et Socie
´te
´’ France. This is a contribution from the Imperial
College Grand Challenges in Ecosystem and the Environment Initiative.
Author Contributions T.N.,L.N.H., S.L.L.H., S.C., I.L.,B.C., D.W.P., R.M.E., G.M.M., J.P.W.S.
and A.P. designed the project and this study; T.N., L.N.H., I.L., R.A.S., L.B., J.P.W.S. and
A.P. performed the analyses; T.N., L.N.H., S.L.L.H., S.C., D.J.B., A.C., B.C., J.D., A.D.P.,
S.E.-L., M.G.,M.L.K.H., T.A., D.J.I., V.K., L.K., D.L.P.C., C.D.M., Y.P., H.R.P.P., A.R., J.S., H.J.W.
and A.P. collatedthe assemblage composition data; T.N., L.N.H.,S.L.L.H., S.C., A.D.P.,I.L.,
H.R.P.P., J.P.W.S. and A.P. designed the data-collection protocols and database; R.A.S.,
S.D., M.J.E., A.F., Y.I., J.K., M.K., S.M. and E.W. made substantial contributions to the trait
data used in theanalyses and S.L.T. to thesite-level environmental data; R.A.S., A.F.,Y.I.,
S.M., and M.N. generated the maps of species richness used in the model projections;
T.N., L.N.H. and A.P. wrote the manuscript with contributions from G.M.M., L.B., D.W.P.,
R.M.E., A.D.P., H.R.P.P., S.L.L.H., R.A.S., B.C., S.D., A.F., Y.I., J.K., M.K., S.M., J.P.W.S and
S.L.T.; T.N. and L.N.H. contributed equally to the study.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of the paper.
Correspondence and requests for materials should be addressed to
T.N. (Tim.Newbold@unep-wcmc.org).
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved
50|NATURE|VOL520|2APRIL2015
METHODS
Data collation. Between March 2012 and April 2014 we collated among-site com-
parisonsof ecological assemblage composition from published studies(or from un-
publisheddatasets where the methods have been published) examining the effect of
human activitieson more than one named taxon. A full descriptionof how the data
set was assembled and curated is presented elsewhere
16
. We define sites to be in the
same study if they were sampled using the same methodology and the data were
reported in the same paper; therefore, some publications contain multiple studies.
After six monthsof broad searching, we targeted effortstowards under-represented
taxa, habitat types, biomes and regions. We accepted data only from published or
in-press papers, or data collected usinga published methodology, and we required
that the data providers agreed to our making their data publicly available at the
end of the PREDICTS (Projecting Responses of Ecological Diversity In Changing
Terrestrial Systems) project in 2015. We accepted data only where abundance,
occurrence or richness had been measured at two or more samplinglocations and/
or times, and where all sites were sampled using the same sampling procedure and
with either the same effort or site-specific data on effort. We used geographical
coordinates preferentially from the paper or supplied by data providers; but where
coordinates were not thus available, we geo-referenced themfrom maps in the papers.
The final data set came from 378 studies
48–329
and two unpublished datasets (M. E.
Hanley, 2005 and 2011) that were collected with published methods
146
.
Studies compared from 2 to 360 sites (median 515; 82% had $5 sites); most
sampledspecies from multiplefamilies but fewer thanhalf sampled multipleorders.
Over 70% of sites were from studies that sampled entire communities within a
taxonomic group rather than a target list of species. Removing studies having a
target list did not substantially alter model coefficients (results not shown) and
increased the projected global net average loss of local species richness until 2005
by 0.6%. Sites varied in the maximum linear extent sampled (median 106 m; inter-
quartile range 50 m to 354 m). Model coefficients for the approximately 50% of
studies that reported maximum linear extent were robust to its inclusion in the
models (results not shown).
The great majority of listed taxa were species level, although many could not be
given explicit species identifications (for example, morpho-species)
16
; henceforth
we refer to distinct taxa in our data set as species. We matched taxonomic names
given in the source paper to the Catalogue of Life 2013 Annual Checklist (COL)
330
,
obtaining the full taxonomic classification. In order to relatethe taxonomic names
to species-level trait databases, we generated, for each taxon, a ‘best-guess’ Latin
binomial as: (a) the taxon name from COL if the COL query returned a species-
level identification; (b)the first two words of the text returned by the COL query if
this was a sub-speciesdesignation; (c) the first two words of the taxon name in the
source publication if the COL query returned neither a speciesor sub-species name,
and the taxon name in the source publication contained two or more words. Taxa
that met none of these criteria were not matched to trait data, but were included in
the calculation of richness and total abundance, and for estimating turnover in
community composition among sites.
The resulting data set containeddata for 26,953 species at 11,525 sites. For many
high-diversity taxa, thedatabase contains data for more than 1% of the number of
species thought to have been formally described (Extended Data Fig. 1a). The dis-
tribution of sites among major biomes is roughly proportional to the amount of
terrestrialnet primary productivity (NPP) fixed within each biome (ExtendedData
Fig. 1b).
Site-level composition and diversity. We computed four site-level biodiversity
metrics: within-sample species richness, total abundance, rarefaction-basedrichness
and community-weighted mean organism size. These were calculated as follows.
Within-sample species richness was calculated as the number of differently-named
taxa recorded ata given site in a standardised samplingunit (a measure also known
as species density
331
). We gave precedence to the author’s classification of species,
even where a search of global databases revealed potential synonymies, because
only certain taxonomic groups could be reliably matched to accepted taxonomies.
This measure of richness is appropriate for conservation questions but among-
site differences could be due to effects on numbers of individuals as well as to
changes in the shapeof the species accumulation curve
331
. We therefore also calcu-
lated rarefaction-based species richness by taking 1,000 random samples of nin-
dividuals from each site, wherenis the smallest totalnumber of individuals recorded
at any sitewithin its study,and calculatingthe mean speciesrichness acrosssamples.
This index could only be calculated for sites where, in addition to the criteria above
being met, abundance was recorded as number of individuals. Rarefied species
richness was rounded to the nearest integer for analysis with Poisson errors.
Total abundance was calculated as the sum of the measures of abundance of all
taxa at a site; we were thus unable to estimateabundance for siteswhere only species
occurrence or overall richness or diversity had been recorded (17% of sites). Some
abundance metrics—those not reported as densities per unit time, distance, area
or volume sampled—were sensitive to sampling effort. When a study reported any
of these metrics and sampling effortvaried among sites withina study, we corrected
the raw abundance measurements for the sampling effort expended at each sam-
pling locationand time. This was done by rescaling the sampling effortswithin each
study so that the most heavily sampled site had a value of one (to prevent intro-
ducing additional heterogeneity into the modelled values), and then dividing the
raw abundance measurements by this relative sampling effort.
Community-weighted meanorganism size wascalculated as the arithmetic mean
of log-transformed heightof plants (availablefor 4,235 speciesin our data set) or the
log-transformed body mass or volume of vertebrates, beetles and hoverflies (5,236
species) present at a site, weighted by abundance
332
. Plant height data were taken
from the TRY database
333
; for 61 species where plant vegetative height data were
unavailable, we estimated it from generative height from a regression across the
2,554 specieswith estimatesof both traits (R
2
50.91). Dataon vertebrate body mass
were taken from the PanTHERIA database for mammals
334
, from BirdLife Inter-
national’s World Bird Database for birds, and from a wide range of published and
grey-literature sources for amphibians
335–381
. Length data for reptiles were take n from
published
382,383
and unpublished (S. Meiri and A. Feldman, unpublished data) sources,
and converted to estimates of body mass using published length-mass allome-
tries
384,385
. Arthropod size data (beetles and hoverflies) were collated from pub-
lished sources
386,387
. Beetle length and amphibian snout-vent length values were
raised to the power three so that they had the same dimensionality as the other
animal size measures. For both plant height and vertebrate body mass, missing
values were interpolated as the average values for congeners, since both of these
traits are strongly conserved phylogenetically (Pagel’s l50.98, 0.997, 0.93, 0.89
for plant height, vertebrate body mass, beetle body length and hoverfly thorax vol-
ume, respectively).
Human pressure data. While manyhuman pressures can affect local biodiversity,
we focus on those that can be obtained for sites aroundthe world and for which, as
far as possible, spatiotemporal data are available for 1500–2095; this focus enables
us to useour statistical models as a basis forprojecting responses through time.Each
site was assigned to one of eight land-use classes based on the description of the
habitat given in the source paper (see Extended Data Table 1 for definitions): pri-
mary vegetation, secondary vegetation (subdivided into mature, intermediate or
young secondary vegetation), plantation forest, cropland, pasture and urban
16
.These
classes were selected to match the land-use classification adopted in the Intergov-
ernmental Panel on Climate Change Representative Concentration Pathways sce-
narios
26
to facilitate the projection of our models onto these scenarios. Sites were
also assigned to a level of human intensity of use (minimal, light or intense) within
each major land-use class, also based on the description of the habitat in the source
paper (see Extended Data Table 1 for definitions). The factors that determined
this level depended on the land-use class (for example, bushmeat extraction and
limited logging in primary and secondary vegetation, or stocking density and
chemical inputs in pasture; Extended Data Table 1). Sites that could not be clas-
sified for land-use and use intensity were excluded from the analyses. The final
dataset contained the following numbers of sites in each land use and land-use
intensity level:primary vegetation, minimal use, 1,546 (from 183 studies);light use,
860 (76 studies); intense use, 449 (33 studies); mature secondary vegetation, min-
imal use, 198 (52 studies); light/intense use, 213 (23 studies); intermediate second-
ary vegetation, minimal use, 404 (55 studies); light/intense use, 269 (30 studies);
young secondary vegetation, minimal use, 431 (50 studies); light/intense use, 331
(34 studies); plantation forest, minimal use, 356 (47 studies); light use, 402 (42
studies); intenseuse, 238 (29 studies); cropland, minimal use, 427 (45 studies); light
use, 632 (43 studies); intense use, 703 (36 studies); pasture, minimal use, 525 (43
studies); light use, 434 (52 studies); intense use, 174 (23 studies); and urban, min-
imal use, 174 (23 studies); light use, 244 (26 studies); intense use, 195 (18 studies).
We overlaid our sites with available globaldata sets to obtain site-level estimates
of human population density
45
, distance to the nearest road
46
and estimated travel
time to nearest population centre with greater than 50,000 inhabitants
47
.Fordis-
tance to nearest road, the map of roads was first projected onto a Berhmann equal-
area projection. These operations were carriedout using Python code implemented
using the arcpy Python module in ArcMap version 10.0 (ref. 388). In the main
figures, the inverses of distanceto roads and travel time to major population centre
(proximity to roads and accessibility) were presented so that high values corre-
sponded to higher hypothesized human effect. To estimate the history of human
use of the landscapes within which sites were located, we calculated the number of
years since the 30-arc-second grid cell containing each site became 30% covered
by human land uses (cropland, pasture and urban), according to the HYDE model
389
.
Co-linearity among variables describing anthropogenic change was low; the highest
correlation was between land use and human population density (Pearson R
2
5
0.31).
Modelling site-level diversity, composition and turnover. The response of site-
level diversity to the measures of anthropogenic change was modelled using gen-
eralized linear mixed effects models, implemented in the lme4 package version
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
1.0-5 (ref. 44) in R version 3.0.2 (ref. 390). We first compared candidate random-
effects structures using the full candidate fixed-effects structure
391
. Random-
intercept terms considered in all models were the identity of the study from which
data were taken,to account for study-level differencesin the response variables and
samplingmethods used, and—within study—thespatial block in which the site was
located, to account for the spatial arrangement of sites. For models of species rich-
ness (within-sample and rarefied),we also fitted an observation-level random effect
(that is, site identity) to account for the overdispersion present
392
. We also con-
sidered random slopes, with respect to study, of each of the main fixed effects (land
use, land-use intensity, human population density, distance to nearest road, travel
time to nearest major city and time since the landscape was majority converted to
human uses). Random effects were retained or discarded based on the models’
Akaike Information Criterion values.
Once the best random-effects structure had been selected, we performed back-
ward stepwise model simplification to select the best fixed-effects structure (see
Supplementary Information)
391
. Human population density, distance to roads, tra-
vel time tonearest major city andtime since major human useof the landscape were
log transformed in the analyses, with a value of 1 added to human population den-
sity, travel time to nearest major city and time since majorlandscape conversion to
deal with zero values. These four variables were fitted as continuous effects, with
quadratic polynomials for human population, distance to roads and travel time to
nearest major city, and as a linear effect for time since human landscape conver-
sion. For variables fitted as quadraticpolynomials, we also tested linear effects dur-
ing the backward stepwise model selection. All continuous variables were rescaled
before analysis so that valuesranged between zero and one. Interaction terms were
tested first, and then removed to test the main effects. All main effects that were
part of significant interaction terms were retained in the final models regardless of
their significance as maineffects. Forthe model of community-weightedmean body
mass and plant height, because the number of sites with data was smaller than for
the other metrics, only land use (excluding urban sites, which were few), human
population density and distance to roads, and no interactions, were fitted (for the
model of plant height, sample sizes in each land use were: primary vegetation, 634
sites; secondary vegetation,851 sites; plantation forest, 222 sites; cropland, 72 sites;
pasture, 412 sites; and for the modelof animal mass: primary vegetation, 1728 sites;
secondary vegetation,805 sites; plantation forest, 602 sites; cropland,641 sites; pas-
ture, 440 sites).The decision whether or not to retain termswas based on likelihood
ratio tests. The coefficient estimates of the best models are shown in Fig. 1b–d and
Extended Data Fig. 2, and theformulae and statistical results are shown in the Sup-
plementary Information. To test for spatial autocorrelation in the residuals of the
final best models, we calculated Moran’s Ivalues and associated Pvalues, sepa-
rately for each study considered in the models, using the spdep package version
0.5-68 (ref. 393) in R; the distribution of Pvalues across studies was used as an
indication of whether spatial autocorrelation was likely to cause a problem. This
revealed that the residuals showed little spatial autocorrelation (Extended Data
Fig. 5). We used cross validation to assess the robustness ofmodel parameter esti-
mates, first based on dividing the studies randomly into ten equal-sized sets and
dropping each set in turn (ExtendedData Fig. 3c), and second based on leaving out
the studies from each biome in turn (Extended Data Fig. 3d).
Publication bias is a potential problem for any large-scale synthesis of data from
many publications. In standard meta-analyses,funnel plots
394
can be used to test for
any relationship between standard error and effect size, as a bias in effect sizes at
high standard error towardsmore positive or more negative effects indicates a likely
effect of publication bias. Creating funnel plots for our data was more complicated
because ours was a site-level analysis of raw diversity estimates rather than a tra-
ditional meta-analysis. Instead we generatedindividual models relating diversityto
land use for each study that sampled at least two sites within each of at least two
land-use types. We focused on land use because: (a) there were a small number of
sites included in mostwithin-study models; and (b) the original studies focused on
effects of land use, not generally on land-use intensity, human population density
or distance to roads, and thus any effect of publication bias would likely be seen in
the land-use coefficients. Funnel plots were generated by plotting, for each land-
use type, the estimated model coefficients against the associated standard errors
(Extended Data Fig. 4). There were some indications of an effect of publication
bias, with less certaincoefficient estimates tending to have more negativeestimates
for some of the land uses (Extended Data Fig. 4). However, study-level random
slopes of human-dominated land uses tended to be more negative for studies that
sampled more sites (Extended Data Fig. 4). It is important to emphasize that in a
site-level analysis likeours, studies with fewer sites have less weight in the models.
Modelled coefficient estimates were generallyrobust to the removal of these studies
(Extended Data Fig. 4). Basing projections on coefficient estimates from models
where small studies were excluded led to a less than 1% change in the estimated
global richness values (results not shown). As with all studies based on data from
the literature, we underrepresent unpublished data.
To model turnover of species composition between pairs of sites, we calculated
average dissimilarity
23
in the lists of present species (1 2Sørensen index) between
all pairs of sites within each study. For this analysis, we were only able to consider
studies with more than one site in at least one of the land-use types considered.
Once compositional similarity had been calculated for every pair of sites within
each study, the average compositional similarity was calculated for every pair of
land-use types consideredwithin each study (including comparisons between sites
in the same land-use type).Finally, the average compositionalsimilarity was calcu-
lated for each pair of land-usetypes across all studies. To visualize the clustering of
different land-usetypes in terms of community composition, we performed a hier-
archical complete-linkage cluster analysis on the compositional dissimilarity (that
is, 1 2similarity) matrix, usingthe hclust function in R version 3.0.2 (ref. 390). To
test whether differences in the average geographicdistance between pairs of sites in
different land-use combinations affected these results, we correlated average com-
positional similarity with average distance between sites, for all pairwise combina-
tions of landuse (including comparisons of a land-use type with itself). Correlations
between average distance and average community similarity were only very weakly
negative (R
2
50.001), suggesting they do not strongly distort the comparisons of
communitycomposition. However, the fact that someland uses tend to occur more
closely together than others could influence the diversity patterns seen in our
models, if some land uses are typically close to high-diversity habitats and so are
more likely to benefit from dispersal. For example, sites in secondary vegetation
and plantation forestwere closer, on average, to primary vegetationsites than were
those in cropland, pasture and urban (averagedistances to sites in primary vegeta-
tion were:other primary vegetation sites 57.38 km;mature secondary vegetation5
4.4 km; intermediate secondaryvegetation53.9 km; young secondaryvegetation5
6.9 km; plantation forest54.2 km; cropland 516.4 km; pasture 510.1 km; and
urban 511.4 km). Accounting for distance in such already complex models is not
computationally tractable. In making the projections, we therefore implicitly assume
that the average distances will not change (that is, that secondary vegetation and
plantation forests will remain closer to primary vegetation than cropland, pasture
and urban habitats).
Projecting the models onto spatial estimates of anthropogenic variables. We
projected the best overall models of richness (within-sample and rarefied), abun-
dance and community-weighted mean organism size onto estimates of land use,
land-use intensity and human population density at 0.5u30.5uresolution, using
historical estimates for 1500 to 2005, and four RCP scenarios of future changes
(IMAGE 2.6, MiniCAM 4.5, AIM 6.0 and MESSAGE 8.5; the names refer to the
integrated assessment models used and the numbers to the amount of radiative
forcing assumed in 2100)
395
. In the absence of global projections, proximity to
roads and accessibility were omitted from our projections.
Estimates of land use for both the historical reconstruction and the future sce-
narios were taken from the harmonized land-use data accompanying the scen-
arios
26
. Estimates of the stage of secondary vegetation (young, intermediate or
mature) are not available directly in the RCP land-use data. However, these data
contain estimates of the transition each year between secondary vegetation and all
other land-use types. To convert this into an estimate of the proportion of second-
ary vegetation in each of the stages of maturity, we considered any transition to
secondary vegetation to result in secondary vegetation of age zero. Each year, this
age was then incremented by one. In the absence of better information, any tran-
sitions from secondary vegetation to any other land-use type were assumed to be
drawn evenly from the ages currently represented. For the purposes of the projec-
tions, secondary vegetation was considered to be young until an age of 30 years,
intermediatebetween 30 years and 100 years, and mature thereafter. We developed
C# code to convert land-use transitions into estimates of the stage structure of
secondary vegetation.
Gridded temporal estimates of human population density were directly avail-
able for the HYDE historical scenario and MESSAGE future scenario. Human
population trajectories in the MiniCAM model were resolved only to the level of
United Nations regions
41
; we therefore downscaled these to grid cells assuming no
temporal changein the spatial pattern of relativepopulation density within regions
compared to present day patterns
45
, which is the method used in other RCP-
scenario land-use models lacking human population data resolved to grid cells
26
.
Gridded estimates of human population from the MESSAGE model were down-
loaded from http://www.iiasa.ac.at/web-apps/ggi/GgiDb/. For the scenarios for which
human population projections were not available (IMAGE and AIM), we used
country-level estimates from the ‘medium’ scenario of the United Nations popu-
lation division
396
, which gives the closest global predictions of future human po-
pulation to those assumed by IMAGE and AIM
26
. These country-level estimates
were downscaled to grid cells using the same method as for MiniCAM’s regional
projections.
Land-use intensity was an important explanatory variable in our models, but
global maps of land-use intensity are not available. We therefore generated global
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved
estimates of current land-use intensity based on a map of ‘Global Land Systems’
397
,
which divides coarseland-use types into sub-categories basedon levels of cropland
intensity, livestock densities and human population density. We mapped each
Global Land Systems class onto one or more relevant combinations of our classes
of land use and land-use intensity (Extended Data Table 2). The Global Land Sys-
tems data set has a spatial resolution of 5 arcmin. To calculate the proportion of
each 0.5u30.5ucelloccupied by each land use and land-use intensity combination
we calculated the proportionof 5-arcmin cells within each 0.5u30.5ucell contain-
ing matching GlobalLand Systems categories (see legend of Extended DataTable 2
for details).
To generate past and future estimates of land-use intensity, we modelled the
current proportion of each land-use type estimated to be under minimal, light or
intense levels of intensity within each grid cell (one model for each intensity level),
as a function of the prevalence of the land-use type within each cell and human
population density, with the relationships allowed to vary among the 23 United
Nations (UN) sub-regions (that is, we fitted interaction terms between UN sub-
region and both the prevalence of each land-use type and human population den-
sity). UN sub-region data were taken from the world borders shapefile version 0.3
(http://thematicmapping.org/downloads/world_borders.php) and converted to a
0.5u30.5uraster using ArcMap version10.0 (ref. 388). The models weredeveloped
using generalized linearmodels with a binomial distributionof errors, implemented
in the lme4package version 1.0-5(ref. 44) in R version 3.0.2 (ref. 390). The resulting
models explained between 30.6% and 76.7% of the deviance in estimated current
levels of intensity. Past and future land-use intensities were estimated by applying
the models to the same past and future estimates of land use and human popu-
lation density as above.
The scenarios gave the proportion of each grid cell estimated to be occupied by
each combinationof land use and land-use intensity.We did not attempt to resolve
human populationdensity within grid cells for our historical estimates or forecasts,
thereby assuming it to be spatially (but not temporally) constant within each cell.
The coefficients from the models of site-level diversity were thus applied to each
combinationof land use and intensity within each cell, with the same human popu-
lation density estimate acrossall combinations. All predictions were expressed as a
percentage net changecompared with a baseline before human land-use effects on
biodiversity, in which all land use was assumed to be primary vegetation of min-
imal intensity of use, and with a human population density of zero. Each cell’s
average value of net biodiversity change was calculated as the area-weighted mean
value across all land uses and intensities. Global average values were calculated as
mean values across all cells, weighted by cell area and an appropriate weighting
factor to account for the fact that cells have different baseline levels of diversity.
The weighting factors applied were: terrestrial vertebrate species richness in the
case of richness, and net primary production (NPP) in the case of total abundance.
No weighting factor was appliedfor projectionsof community-weighted meanplant
height. Terrestrial vertebrate species richness was estimated by overlaying extent-
of-occurrence range maps for mammals, birds, amphibians and reptiles, using
Python code written by ourselves and implemented in ArcMap version 10.0 (ref. 388).
Data on NPP were estimates of potential NPP (that is, in the absence of human
effects) from the Lund-Postdam-Jena (LPJ) Dynamic Global Vegetation Model
398
.
The 95% confidence intervals around the projected values of biodiversity for
each combination of pressure variables were estimated based on uncertainty in the
modelled coefficients. We were unable to conduct multi-model averaging to account
for uncertainty in the structure of the models (that is, projections were based only
on the final best model) because applying such complex mixed-effects models,
based on such large data sets, to multiple scenarios of human pressure at a global
scale was intractable both in terms of time and computer-memory requirements.
We were also unable to account for uncertainty in the trajectories of the human
pressure variables, because uncertainty estimates are not available for any of the
variables considered.
To estimate average biodiversity change in individual countries, we intersected
the gridded projections with the world borders shapefile (see above) using the ex-
tract function in the raster package version 2.2-12 (ref. 399) in R version 3.0.2 (ref. 390).
Mean values acrossthe cells associated with each country were calculated, weighted
by cell area. To interpret the outcomes for countries in terms of their natural bio-
diversity, we related the country-level projections to estimates of average natural
vertebrate species richness (see above). To interpret the outcomes for countries in
terms of their socio-economic status, we related the projections to estimates of the
Human Development Index, which is an indicator of education, life expectancy,
wealth and standard of living (https://data.undp.org/).
48. Aben, J., Dorenbosch, M., Herzog, S. K., Smolders, A. J. P. & Van Der
Velde, G. Human disturbance affects a deciduous forest bird community
in the Andean foothills of central Bolivia. Bird Conserv. Int. 18, 363–380
(2008).
49. Adum, G. B., Eichhorn, M. P., Oduro, W., Ofori-Boateng, C. & Rodel, M. O. Two-
stage recovery of amphibian assemblages following selective logging of tropical
forests. Conserv. Biol. 27, 354–363 (2013).
50. Aguilar Barquero, V. & Jime
´nez Herna
´ndez, F. Diversidad y distribucio
´nde
palmas (Arecaceae) en tres fragmentos de bosque muy hu
´medo en Costa Rica.
Rev. Biol. Trop. 57, 83–92 (2009).
51. Alberta Biodiversity Monitoring Institute (ABMI). The raw soil arthropods dataset
and the raw trees & snags dataset from Prototype Phase (2003-2006) and
Rotation 1 (2007-2012). (2013).
52. Alcala, E. L., Alcala, A. C. & Dolino, C. N. Amphibians and reptiles in tropical
rainforest fragments on Negros Island, the Philippines. Environ. Conserv. 31,
254–261 (2004).
53. Alcayaga, O. E., Pizarro-Araya, J., Alfaro, F. M. & Cepeda-Pizarro, J. Spiders
(Arachnida, Araneae) associated to agroecosystems in the Elqui Valley
(Coquimbo Region, Chile). Revista Colombiana De Entomologia 39, 150–154
(2013).
54. Ancrenaz, M., Goossens, B., Gimenez, O., Sawang, A. & Lackman-Ancrenaz, I.
Determination of ape distribution and population size using ground and aerial
surveys: a case study withorang-utans in lower Kinabatangan, Sabah, Malaysia.
Anim. Conserv. 7, 375–385 (2004).
55. Arbela
´ez-Corte
´s, E., Rodrı
´guez-Correa, H. A. & Restrepo-Chica, M. Mixed bird
flocks: patterns of activity and species composition in a region of the Central
Andes of Colombia. Revista Mexicana De Biodiversidad 82, 639–651 (2011).
56. Armbrecht, I., Perfecto, I. & Silverman, E. Limitation of nesting resources for ants
in Colombian forests and coffee plantations. Ecol.Entomol. 31, 403–410 (2006).
57. Arroyo, J., Iturrondobeitia, J. C., Rad, C. & Gonzalez-Carcedo, S. Oribatid mite
(Acari) community structure in steppic habitats of Burgos Province, central
northern Spain. J. Nat. Hist. 39, 3453–3470 (2005).
58. Azhar, B. et al. The influence of agricultural system, stand structural complexity
and landscape context on foraging birds in oil palm landscapes. Ibis 155,
297–312 (2013).
59. Azpiroz, A. B. & Blake, J. G. Avian assemblages in altered and natural grasslands
in the northern Campos of Uruguay. Condor 111, 21–35 (2009).
60. Baeten, L. et al. Early trajectories of spontaneous vegetation recovery after
intensive agricultural land use. Restor. Ecol. 18, 379–386 (2010).
61. Baeten, L., Hermy, M., Van Daele, S. & Verheyen, K. Unexpected understorey
community development after 30 years in ancient and post-agricultural forests.
J. Ecol. 98, 1447–1453 (2010).
62. Ba
´ldi, A., Bata
´ry, P. & Erdo˝ s, S. Effects of grazing intensity on bird assemblages
and populations of Hungarian grasslands. Agric.Ecosyst. Environ. 108, 251–263
(2005).
63. Banks, J. E., Sandvik,P. & Keesecker, L. Beetle(Coleoptera) and spider (Araneae)
diversity in a mosaic of farmland, edge, and tropical forest habitats in western
Costa Rica. Pan-Pac. Entomol. 83, 152–160 (2007).
64. Barlow, J. et al. Quantifying the biodiversity value of tropical primary, secondary,
and plantation forests. Proc. Natl Acad. Sci. USA 104, 18555–18560 (2007).
65. Bartolommei, P., Mortelliti, A., Pezzo, F. & Puglisi, L. Distribution of nocturnal
birds (Strigiformes and Caprimulgidae) in relation to land-use types, extent and
configuration in agricultural landscapes of Central Italy. Rendiconti Lincei-
Scienze Fisiche E Naturali 24, 13–21 (2013).
66. Basset, Y. et al. Changes in Arthropod assemblages along a wide gradient of
disturbance in Gabon. Conserv. Biol. 22, 1552–1563 (2008).
67. Bates, A. J. et al. Changing bee and hoverfly pollinator assemblages along an
urban-rural gradient. PLoS ONE 6, (2011).
68. Baur, B. et al. Effects of abandonment of subalpine hay meadows on plant and
invertebrate diversity in Transylvania, Romania. Biol. Conserv. 132, 261–273
(2006).
69. Berg, A., Ahrne, K., Ockinger, E., Svensson, R. & Soderstrom, B. Butterfly
distribution and abundance is affected by variation in the Swedish forest-
farmland landscape. Biol. Conserv. 144, 2819–2831 (2011).
70. Bernard, H.,Fjeldsa, J. & Mohamed, M. A case studyon the effects of disturbance
and conversionof tropical lowland rain forest on the non-volant small mammals
in north Borneo: management implications. Mammal Study 34, 85–96 (2009).
71. Berry, N. J. et al. The high value of logged tropical forests: lessons from northern
Borneo. Biodivers. Conserv. 19, 985–997 (2010).
72. Bicknell, J. & Peres, C. A. Vertebrate population responses to reduced-impact
logging in a neotropical forest. For. Ecol. Manage. 259, 2267–2275 (2010).
73. Bihn, J. H., Verhaagh, M., Braendle, M. & Brandl, R. Do secondary forests act as
refuges for old growth forest animals? Recovery of ant diversity in the Atlantic
forest of Brazil. Biol. Conserv. 141, 733–743 (2008).
74. Billeter, R. et al. Indicators for biodiversity in agricultural landscapes: a pan-
European study. J. Appl. Ecol. 45, 141–150 (2008).
75. Bo
´çon, R. Riqueza e abunda
ˆncia de aves em tre
ˆsesta
´gios sucessionais da floresta
ombro
´fila densa submontana, Antonina, Parana
´.PhD thesis, Universidade Federal
do Parana
´(2010).
76. Borges, S. H. Bird assemblages in secondary forests developing after slash-and-
burn agriculture in the Brazilian Amazon. J. Trop. Ecol. 23, 469–477 (2007).
77. Boutin, C., Baril, A. & Martin, P. A. Plant diversity in crop fields and woody
hedgerows of organic and conventional farms in contrasting landscapes. Agric.
Ecosyst. Environ. 123, 185–193 (20 08).
78. Bouyer, J. et al. Identification of ecological indicators for monitoring ecosystem
health in the trans-boundary W Regional park: a pilot study. Biol. Conserv. 138,
73–88 (2007).
79. Bragagnolo,C., Nogueira, A. A.,Pinto-da-Rocha, R. & Pardini, R. Harvestmen inan
Atlantic forest fragmented landscape: evaluating assemblage response to
habitat quality and quantity. Biol. Conserv. 139, 389–400 (2007).
ARTICLE RESEARCH
G2015 Macmillan Publishers Limited. All rights reserved
80. Brearley, F. Q. Below-ground secondary successionin tropical forestsof Borneo.
J. Trop. Ecol. 27, 413–420 (2011).
81. Brito, I., Goss, M. J., de Carvalho, M., Chatagnier, O. & van Tuinen, D. Impact of
tillage system on arbuscular mycorrhiza fungal communities in the soil under
Mediterranean conditions. Soil Tillage Res. 121, 63–67 (2012).
82. Brunet, J. et al. Understory succession in post-agricultural oak forests: habitat
fragmentation affects forest specialists and generalists differently. For. Ecol.
Manage. 262, 1863–1871 (2011).
83. Buczkowski, G. Extreme life history plasticity and the evolution of invasive
characteristics in a native ant. Biol. Invasions 12, 3343–3349 (2010).
84. Buczkowski, G. & Richmond, D. S. The effect of urbanization on ant abundance
and diversity:a temporal examinationof factors affecting biodiversity. PLoS ONE
7, (2012).
85. Buddle, C. M. & Shorthouse, D. P. Effects of experimental harvesting on spider
(Araneae)assemblages in boreal deciduousforests. Can. Entomol.140, 437–452
(2008).
86. Buscardo, E. et al. The early effects of afforestation on biodiversity of grasslands
in Ireland. Biodivers. Conserv. 17, 1057–1072 (2008).
87. Cabra-Garcı
´a, J., Bermu
´dez-Rivas, C., Osorio, A. M. & Chaco
´n, P. Cross-taxon
congruenceof alpha and beta diversityamong five leaf litter arthropodgroups in
Colombia. Biodivers. Conserv. 21, 1493–1508 (2012).
88. Ca
´ceres, N. C., Napoli, R. P., Casella, J. & Hannibal, W. Mammals in a fragmented
savannah landscape in south-western Brazil. J. Nat. Hist. 44, 491–512 (2010).
89. Cagle, N. L. Snake species distributions and temperate grasslands: a case study
from the American tallgrass prairie. Biol. Conserv. 141, 744–755 (2008).
90. Calvin
˜o-Cancela,M., Rubido-Bara
´, M. & van Etten,E. J. B. Do eucalypt plantations
provide habitat for native forest biodiversity? For. Ecol. Manage. 270, 153–162
(2012).
91. Cameron, S. A. et al. Patterns of widespread decline in North American bumble
bees. Proc. Natl Acad. Sci. USA 108, 662–667 (2011).
92. Carrijo, T. F., Brandao, D., de Oliveira, D. E., Costa, D. A. & Santos, T. Effects of
pasture implantation on the termite (Isoptera) fauna in the Central Brazilian
Savanna (Cerrado). J. Insect Conserv. 13, 575–581 (2009).
93. Carvalho, A. L. d., Ferreira, E. J. L., Lima, J. M. T. & de Carvalho, A. L. Floristic and
structural comparisons among palm communities in primary and secondary
forest fragments of the Raimundo Irineu Serra Environmental Protection Area -
Rio Branco, Acre, Brazil. Acta Amazon. 40, 657–666 (2010).
94. Castro, H., Lehsten, V., Lavorel, S. & Freitas, H. Functional response traits in
relationto land use change in the Montado. Agric. Ecosyst. Environ.137, 183–191
(2010).
95. Castro-Luna,A. A., Sosa, V. J. & Castillo-Campos, G. Bat diversity and abundance
associatedwith the degree of secondary succession in a tropicalforest mosaic in
south-eastern Mexico. Anim. Conserv. 10, 219–228 (2007).
96. Center ForInternational ForestryResearch (CIFOR).MultidisciplinaryLandscape
Assessment — Cameroon. http://www.cifor.org/mla (2013).
97. Center ForInternational ForestryResearch (CIFOR).MultidisciplinaryLandscape
Assessment — Philippines. http://www.cifor.org/mla (2013).
98. Centro Agrono
´mico Tropical de Investigacio
´n y Ensen
˜anza(CATIE). Unpublished
data of reptilian and amphibian diversity in six countries in Central America (Centro
Agrono
´mico Tropical de Investigacio
´n y Ensen
˜anza (CATIE), 2010).
99. Cerezo, A., Conde, M. & Poggio, S. Pasturearea and landscape heterogeneity are
key determinants of bird diversity in intensively managed farmland. Biodivers.
Conserv. 20, 2649–2667 (2011).
100. Chapman, K. & Reich, P. Land use and habitat gradients determine bird
community diversity and abundance insuburban, rural and reservelandscapes
of Minnesota, USA. Biol. Conserv. 135, 527–541 (2007).
101. Chauvat, M., Wolters, V. & Dauber, J. Response of collembolan communities to
land-use change and grassland succession. Ecography 30, 183–192 (2007).
102. Clarke, F. M., Rostant, L. V. & Racey, P. A. Life after logging: post-logging recovery
of a neotropical bat community. J. Appl. Ecol. 42, 409–420 (2005).
103. Cleary, D. F. R. et al. Diversity and community composition of butterflies and
odonates in an ENSO-induced fire affected habitat mosaic: a case study from
East Kalimantan, Indonesia. Oikos 105, 426–448 (2004).
104. Cleary, D. F. R. & Mooers, A. O. Burning and logging differentially affect endemic
vs. widely distributed butterfly species in Borneo. Divers. Distrib. 12, 409–416
(2006).
105. Cockle, K. L., Leonard,M. L. & Bodrati, A. A. Presence andabundance of birds in an
Atlantic forest reserve and adjacent plantation of shade-grown yerba mate, in
Paraguay. Biodivers. Conserv. 14, 3265–3288 (2005).
106. Connop, S., Hill, T., Steer, J. & Shaw, P. Microsatellite analysis reveals the spatial
dynamics of Bombus humilis and Bombus sylvarum.Insect Conserv. Divers. 4,
212–221 (2011).
107. D’Aniello, B., Stanislao, I., Bonelli, S. & Balletto, E. Haying and grazing effects on
the butterfly communities of two Mediterranean-area grasslands. Biodivers.
Conserv. 20, 1731–1744 (2011).
108. Darvill, B., Knight, M. E. & Goulson, D. Use of genetic markers to quantify
bumblebee foraging range and nest density. Oikos 107, 471–478 (2004).
109. Davis, A. L. V. & Philips, T. K. Effect of deforestation on a southwest Ghana dung
beetle assemblage (Coleoptera: Scarabaeidae) at the periphery of Ankasa
conservation area. Environ. Entomol. 34, 1081–1088 (2005).
110. Davis, E. S., Murray, T. E., Fitzpatrick, U., Brown,M. J. F. & Paxton, R. J. Landscape
effects on extremely fragmented populations of a rare solitary bee, Colletes
floralis.Mol. Ecol. 19, 4922–4935 (2010).
111. Dawson, J. et al. Bird communities of the lower Waria Valley, Morobe Province,
Papua New Guinea: a comparison between habitat types. Trop. Conserv. Sci. 4,
317–348 (2011).
112. Delabie, J. H. C. et al. Ants as biological indicators ofWayana Amerindian landuse
in French Guiana. C. R. Biol. 332, 673–684 (2009).
113. Dieko
¨tter, T., Walther-Hellwig, K., Conradi, M., Suter, M. & Frankl, R. Effects of
landscape elements on the distribution of the rare bumblebee species Bombus
muscorum in an agricultural landscape. Biodivers. Conserv. 15, 57–68 (2006).
114. Domı
´nguez, E., Bahamonde, N. & Mun
˜oz-Escobar, C. Efectos de la extraccio
´nde
turba sobre la composicio
´n y estructurade una turbera de Sphagnum explotada
y abandonada hace 20 an
˜os, Chile. Anales Instituto Patagonia (Chile) 40, 37–45
(2012).
115. Dominguez-Haydar, Y. & Armbrecht, I. Response of ants and their seed removal
in rehabilitation areas and forests at El Cerrejon coal mine in Colombia. Restor.
Ecol. 19, 178–184 (2011).
116. Dumont, B. et al. How does grazingintensity influence the diversity of plants and
insects in a species-rich upland grassland on basalt soils? Grass Forage Sci. 64,
92–105 (2009).
117. Dures, S. G. & Cumming, G. S. The confounding influence of homogenising
invasivespecies in a globally endangeredand largely urban biome: Doeshabitat
quality dominate avian biodiversity? Biol. Conserv. 143, 768–777 (2010).
118. Edenius, L., Mikusinski, G. & Bergh, J. Can repeated fertilizer applications to
young Norway spruce enhance avian diversity in intensively managed forests?
Ambio 40, 521–527 (2011).
119. Elek, Z. & Lovei, G. L. Patterns in ground beetle (Coleoptera: Carabidae)
assemblages along an urbanisation gradient in Denmark. Acta Oecologica 32,
104–111 (2007).
120. Endo, W. et al. Game vertebrate densities in hunted and nonhunted forest sites in
Manu National Park, Peru. Biotropica 42, 251–261 (2010).
121. Faruk, A., Belabut, D., Ahmad, N., Knell, R. J. & Garner, T. W. J. Effects of oil-palm
plantations on diversity of tropical anurans. Conserv. Biol. 27, 615–624 (2013).
122. Farwig, N., Sajita, N. & Boehning-Gaese, K. Conservation value of forest
plantations for bird communities in western Kenya. For. Ecol. Manage. 255,
3885–3892 (2008).
123. Fayle, T. M. et al. Oil palm expansion into rain forest greatly reduces ant
biodiversity in canopy, epiphytes and leaf-litter. Basic Appl. Ecol. 11, 337–345
(2010).
124. Felton, A. M., Engstrom, L. M., Felton, A. & Knott, C. D. Orangutan population
density, forest structure and fruit availability in hand-logged and unlogged peat
swamp forests in West Kalimantan, Indonesia. Biol. Conserv. 114, 91–101
(2003).
125. Fensham, R., Dwyer, J., Eyre, T., Fairfax, R. & Wang, J. The effect of clearing on
plant composition in mulga (Acacia aneura) dryforest, Australia. AustralEcol. 37,
183–192 (2012).
126. Fermon, H., Waltert, M., Vane-Wright, R. I. & Muhlenberg, M. Forest use and
vertical stratification in fruit-feeding butterflies of Sulawesi, Indonesia: impacts
for conservation. Biodivers. Conserv. 14, 333–350 (2005).
127. Ferreira, C. & Alves, P. C. Impacto da implementaça
˜o de medidas de gesta
˜odo
habitat naspopulaço
˜es de coelho-bravo(Oryctolagus cuniculus algirus) no Parque
Natural do Sudoeste Alentejano e Costa Vicentina. (Centro de Investigaça
˜oem
Biodiversidade e Recursos Gene
´ticos (CIBIO, 2005).
128. Fierro, M. M., Cruz-Lopez, L., Sanchez, D., Villanueva-Gutierrez,R. & Vandame, R.
Effect ofbiotic factors on the spatialdistribution of stingless bees (Hymenoptera:
Apidae, Meliponini) in fragmented neotropical habitats. Neotrop. Entomol. 41,
95–104 (2012).
129. Filgueiras, B., Iannuzzi, L. & Leal, I. Habitat fragmentation alters the structure of
dung beetle communities in the Atlantic Forest. Biol. Conserv. 144, 362–369
(2011).
130. Flaspohler, D. J. et al. Long-term effects of fragmentation and fragment
properties on bird species richness in Hawaiian forests. Biol. Conserv. 143,
280–288 (2010).
131. Fukuda, D., Tisen, O. B., Momose, K. & Sakai, S. Bat diversity in the vegetation
mosaic around a lowland dipterocarp forest of Borneo. Raffles Bull. Zool. 57,
213–221 (2009).
132. Furlani, D., Ficetola, G. F., Colombo, G., Ugurlucan, M. & De Bernardi, F.
Deforestation and the structure of frog communities in the Humedale Terraba-
Sierpe, Costa Rica. Zoolog. Sci. 26, 197–202 (2009).
133. Garden, J. G., McAlpine, C. A. & Possingham, H. P. Multi-scaled habitat
considerations for conserving urban biodiversity: native reptiles and small
mammals in Brisbane, Australia. Landscape Ecol. 25, 1013–1028 (2010).
134. Gardner, T. A., Hernandez, M. I. M., Barlow, J. & Peres, C. A. Understanding the
biodiversity consequences of habitat change: the value of secondary and
plantationforests for neotropicaldung beetles. J. Appl.Ecol. 45, 883–893 (2008).
135. Gheler-Costa,C., Vettorazzi, C. A., Pardini,R. & Verdade, L. M. The distributionand
abundance of small mammals in agroecosystems of southeastern Brazil.
Mammalia 76, 185–191 (2012).
136. Giordani, P. Assessing the effects of forest management on epiphytic lichens in
coppiced forests using different indicators. Plant Biosyst. 146, 628–637 (2012).
137. Giordano, S. et al. Biodiversity and trace element contentof epiphytic bryophytes
in urban and extraurban sites of southern Italy. Plant Ecol. 170, 1–14 (2004).
138. Golodets, C., Kigel, J. & Sternberg,M. Recovery of plant species composition and
ecosystem function after cessation of grazing in a Mediterranean grassland.
Plant Soil 329, 365–378 (2010).
139. Gottschalk, M. S., De Toni, D. C., Valente, V. L. S. & Hofmann, P. R. P. Changes in
Brazilian Drosophilidae (Diptera) assemblages across an urbanisation gradient.
Neotrop. Entomol. 36, 848–862 (2007).
140. Goulson, D. et al. Effects of land use at a landscape scale on bumblebee nest
density and survival. J. Appl. Ecol. 47, 1207–1215 (2010).
141. Goulson, D., Lye, G. C. & Darvill, B. Diet breadth, coexistence and rarity in
bumblebees. Biodivers. Conserv. 17, 3269–3288 (2008).
RESEARCH ARTICLE
G2015 Macmillan Publishers Limited. All rights reserved
142. Gove, A. D., Majer, J. D. & Rico-Gray, V. Methodsfor conservation outsideof formal
reserve systems: the case of ants in the seasonally dry tropics of Veracruz,
Mexico. Biol. Conserv. 126, 328–338 (2005).
143. Grogan, J. et al. What loggers leave behind: impacts on big-leaf mahogany
(Swietenia macrophylla) commercial populations and potential for post-logging
recovery in the Brazilian Amazon. For. Ecol. Manage. 255, 269–281 (2008).
144. Gu, W.-B., Zhen-Rong,Y. & Dun-Xiao, H. Carabidcommunity and its fluctuation in
farmland of salinity transforming area in the North China Plain: a case study in
Quzhou County, Hebei Province. Biodivers. Sci. 12, 262–268 (2004).
145. Gutierrez-Lamus, D. L. Composition and abundance of Anura in two forest types
(natural and planted) in the eastern cordillera of Colombia. Caldasia 26,
245–264 (2004).
146. Hanley, M. E. et al. Increasedbumblebee abundance alongthe margins of a mass
flowering crop: evidence for pollinator spill-over. Oikos 120, 1618–1624 (2011).
147. Hanson, T. R., Brunsfeld, S. J., Finegan, B. & Waits, L. P. Pollen dispersal and
genetic structure of the tropical tree Dipteryx panamensis in a fragmented Costa
Rican landscape. Mol. Ecol. 17, 2060–2073 (2008).
148. Hashim, N., Akmal, W., Jusoh, W. & Nasir, M. Ant diversity in a Peninsular
Malaysian mangrove forest and oil palm plantation. Asian Myrmecology 3, 5–8
(2010).
149. Hatfield, R. G. & LeBuhn, G. Patch and landscape factors shape community
assemblage of bumble bees, Bombus spp. (Hymenoptera: Apidae), in montane
meadows. Biol. Conserv. 139, 150–158 (2007).
150. Hawes, J. et al. Diversity and composition of Amazonian moths in primary,
secondary and plantation forests. J. Trop. Ecol. 25, 281–300 (2009).
151. Helden, A. J. & Leather, S. R. Biodiversity on urban roundabouts—Hemiptera,
management and the species-area relationship. Basic Appl. Ecol. 5, 367–377
(2004).
152. Herna
´ndez, L., Delgado, L., Meier, W. & Duran, C. Empobrecimiento de bosques
fragmentados en el norte de la Gran Sabana, Venezuela. Interciencia 37,
891–898 (2012).
153. Herrmann, F., Westphal,C., Moritz, R. F. A. & Steffan-Dewenter, I. Geneticdiversity
and mass resources promote colony size and forager densities of a social bee
(Bombus pascuorum) in agricultural landscape s. Mol. Ecol. 16, 1167–1178
(2007).
154. Hietz, P. Conservation of vascular epiphyte diversity in Mexican coffee
plantations. Conserv. Biol. 19, 391–399 (2005).
155. Higuera, D. & Wolf, J. H. D. Vascularepiphytes in dry oakforests show resilienceto
anthropogenic disturbance,Cordillera Oriental,Colombia. Caldasia 32, 161–174
(2010).
156. Hilje, B. & Aide, T. M. Recovery of amphibianspecies richness and composition in
a chronosequence of secondary forests, northeastern Costa Rica. Biol. Conserv.
146, 170–176 (2012).
157. Hoffmann, A. & Zeller, U. Influence of variations in land use intensity on species
diversityand abundance of small mammalsin the Nama Karoo, Namibia.Belg. J.
Zool. 135, 91–96 (2005).
158. Horgan, F. G. Invasion and retreat: shifting assemblages of dung beetles amidst
changing agricultural landscapes in central Peru. Biodivers. Conserv. 18,
3519–3541 (2009).
159. Hu, C. & Cao, Z. P. Nematode community structure undercompost and chemical
fertilizermanagement practice, in the northChina plain. Exp. Agric. 44, 485–496
(2008).
160. Hylander, K. & Weibull, H. Do time-lagged extinctions and colonizations change
the interpretation of buffer strip effectiveness? – a study of riparian bryophytesin
the first decade after logging. J. Appl. Ecol. 49, 1316–1324 (2012).
161. Hylander, K. & Nemomissa, S. Complementary roles of home gardens and exotic
tree plantations as alternative habitats for plants of the Ethiopian montane
rainforest. Conserv. Biol. 23, 400–409 (2009).
162. Ims, R. A. & Henden, J. A. Collapse of an arctic bird community resulting from
ungulate-induced loss of erect shrubs. Biol. Conserv. 149, 2–5 (2012).
163. Cubides, P. J. I. & Cardona, J. N. U. Anthropogenic disturbance and edge effects
on anuran assemblages inhabiting cloudforest fragments in Colombia. Natureza
& Conservacao 9, 39–46 (2011).
164. Ishitani, M., Kotze, D. J. & Niemela, J. Changes in carabid beetle assemblages
across an urban-rural gradient in Japan. Ecography 26, 481–489 (2003).
165. Jacobs, C. T., Scholtz, C. H., Escobar, F. & Davis, A. L. V. How might intensification
of farminginfluence dung beetle diversity(Coleoptera: Scarabaeidae) in Maputo
Special Reserve (Mozambique)? J. Insect Conserv. 14, 389–399 (2010).
166. Johnson, M. F., Go
´mez, A. & Pinedo-Vasquez, M. Land use and mosquito diversity
in the Peruvian Amazon. J. Med. Entomol. 45, 1023–1030 (2008).
167. Jonsell, M. Old park trees as habitat for saproxylic beetle species. Biodivers.
Conserv. 21, 619–642 (2012).
168. Julier, H. E. & Roulston, T. H. Wild bee abundance and pollination service in
cultivated pumpkins: farm management, nesting behavior and landscape
effects. J. Econ. Entomol. 102, 563–573 (2009).
169. Jung, T. S. & Powell, T. Spatial distribution of meadow jumping mice (Zapus
hudsonius) in logged boreal forest of northwestern Canada. Mamm. Biol. 76,
678–682 (2011).
170. Kapoor, V. Effects of rainforest fragmentation and shade-coffee plantations on
spider communities in the Western Ghats, India. J. Insect Conserv. 12, 53–68
(2008).
171. Kappes, H., Katzschner, L. & Nowak,C. Urban summer heat load: meteorological
data as a proxy for metropolitan biodiversity. Meteorologische Zeitschrift 21,
525–528 (2012).
172. Kati, V., Zografou, K., Tzirkalli, E., Chitos, T. & Willemse, L. Butterfly and
grasshopper diversity patterns in humid Mediterranean grasslands: the roles of
disturbance and environmental factors. J. Insect Conserv. 16, 807–818 (2012).
173. Katovai, E., Burley,A. L. & Mayfield,M. M. Understory plant speciesand functional
diversity in the degraded wet tropical forests of Kolombangara Island, Solomon
Islands. Biol. Conserv. 145, 214–224 (2012).
174. Kessler, M. et al. Tree diversity in primary forestand different land use systemsin
Central Sulawesi, Indonesia. Biodivers. Conserv. 14, 547–560 (2005).
175. Kessler, M. et al. Alpha and beta diversity of plants and animals along a tropical
land-use gradient. Ecol. Appl. 19, 2142–2156 (2009).
176. Knight, M. E. et al. Bumblebee nest density and the scale of available forage in
arable landscapes. Insect Conserv. Divers. 2, 116–124 (2009).
177. Knop, E., Ward, P. I. & Wich, S. A. A comparison of orang-utan density in a logged
and unlogged forest on Sumatra. Biol. Conserv. 120, 183–188 (2004).
178. Kohler, F., Verhulst, J., van Klink, R. & Kleijn, D. At what spatial scale do high-
quality habitats enhance the diversity of forbs and pollinators in intensively
farmed landscapes? J. Appl. Ecol. 45, 753–762 (2008).
179. Koivula, M., Hyyrylainen, V. & Soininen, E. Carabid beetles (Coleoptera:
Carabidae) at forest-farmland edges in southern Finland. J. Insect Conserv. 8,
297–309 (2004).
180. Kolb, A. & Diekmann, M. Effects of environment, habitat configuration and forest
continuity on the distribution of forest plant species. J. Veg. Sci. 15, 199–208
(2004).
181. Ko˝ro
¨si, A
´., Bata
´ry, P., Orosz, A., Re
´dei, D. & Ba
´ldi, A. Effects of grazing, vegetation
structure and landscape complexity on grassland leafhoppers (Hemiptera:
Auchenorrhyncha) and true bugs (Hemiptera: Heteroptera) in Hungary. Insect
Conserv. Divers. 5, 57–66 (2012).
182. Krauss, J., Klein, A. M., Steffan-Dewenter, I. & Tscharntke,T. Effects of habitat area,
isolation, and landscape diversity on plant species richness of calcareous
grasslands. Biodivers. Conserv. 13, 1427–1439 (2004).
183. Krauss, J., Steffan-Dewenter, I. & Tscharntke, T. How does landscape context
contributeto effects of habitatfragmentation on diversityand population density
of butterflies? J. Biogeogr. 30, 889–900 (2003).
184. Kumar, R. & Shahabuddin, G. Effects of biomass extraction on vegetation
structure, diversity and composition of forests in Sariska Tiger Reserve, India.
Environ. Conserv. 32, 248–259 (20 05).
185. Lachat, T. et al. Arthropod diversity in Lama forest reserve (South Benin), a
mosaic of natural, degraded and plantation forests. Biodivers. Conserv. 15, 3–23
(2006).
186. Lantschner, M. V., Rusch, V. & Hayes, J. P. Habitat use by carnivores at different
spatial scales in a plantation forest landscape in Patagonia, Argentina. For. Ecol.
Manage. 269, 271–278 (2012).
187. Lantschner, M. V., Rusch, V. & Peyrou, C. Bird assemblages in pine plantations
replacing native ecosystems in NW Patagonia. Biodivers. Conserv. 17, 969–989
(2008).
188. Latta, S. C., Tinoco, B. A., Astudillo, P. X. & Graham, C. H. Patterns and magnitude
of temporal changein avian communities in the Ecuadorian Andes. Condor 113,
24–40 (2011).
189. Le
´gare
´,J.-P.,He
´bert, C. & Ruel, J.-C. Alternative silviculturalpractices in irregular
boreal forests: response of beetle assemblages. Silva Fennica 45, 937–956
(2011).
190. Letcher, S. G. & Chazdon, R. L. Rapid recovery of biomass, species richness, and
species composition in a forest chronosequence in northeastern Costa Rica.
Biotropica 41, 608–617 (2009).
191. Littlewood, N. A., Pakeman, R. J. & Pozsgai, G. Grazing impacts on
Auchenorrhyncha diversity and abundance on a Scottish upland estate. Insect
Conserv. Divers. 5, 67–74 (2012).
192. Liu, Y. H., Axmacher, J. C., Wang, C. L., Li, L. T. & Yu, Z. R. Ground beetle
(Coleoptera: Carabidae) assemblages of restored semi-natural habitats and
intensivelycultivated fields in northern China. Restor. Ecol. 20, 234–239 (2012).
193. Lo-Man-Hung, N. F., Gardner, T. A., Ribeiro-Ju
´nior, M. A., Barlow, J. & Bonaldo,
A. B. The value of primary, secondary, and plantation forests for Neotropical
epigeic arachnids. J. Arachnol. 36, 394–401 (2008).
194. Lo
´pez-Quintero, C. A., Straatsma, G., Franco-Molano, A. E. & Boekhout, T.
Macrofungal diversity in Colombian Amazon forests varies with regions and
regimes of disturbance. Biodivers. Conserv. 21, 2221–2243 (2012).
195. Louhaichi, M., Salkini, A. K. & Petersen, S. L. Effect of small ruminantgrazing on
the plant communitycharacteristics of semiaridMediterranean ecosystems.Int.
J. Agric. Bio. 11, 681–689 (2009).
196. Lucas-Borja, M. E. et al. The effects of human trampling on the microbiological
propertiesof soil and vegetation in Mediterraneanmountain areas. Land Degrad.
Dev. 22, 383–394 (2011).
197. Luja, V., Herrando-Perez, S., Gonzalez-Solis, D. & Luiselli, L. Secondary rain
forests are not havens for reptile species in tropical Mexico. Biotropica 40,
747–757 (2008).
198. Luskin, M. S. Flying