ArticlePDF Available
Contributed Paper
Biodiversity loss along a gradient of deforestation
in Amazonian agricultural landscapes
Thibaud Deca¨
ens ,1Marl´
ucia B. Martins,2Alexander Feijoo,3Johan Oszwald,4Sylvain Dol´
edec,5
J´
erˆ
ome Mathieu,6Xavier Arnaud de Sartre,7Diego Bonilla,8George G. Brown,9
Yeimmy Andrea Cuellar Criollo,10 Florence Dubs,11 Ivaneide S. Furtado,2Val´
erie Gond,12
Erika Gordillo,10 SolenLeClech,
4,13 Rapha¨
el Marichal,14 Danielle Mitja,15 Izildinha Miranda de
Souza,16 Catarina Praxedes,2Rodolphe Rougerie,17 Dar´
ıo H. Ruiz,3Joel Tupac Otero,18
Catalina Sanabria,19 Alex Velasquez,10 Luz Elena M. Zararte,20 and Patrick Lavelle21
1Centre d’Ecologie Fonctionnelle et Evolutive, UMR 5175, CNRS–Univ Montpellier–Univ Paul-Val´
ery–EPHE–SupAgro Montpellier–
INRA–IRD, Montpellier, France
2Laborat´
orio de Ecologia de Invertebrados, Coordenac¸˜
ao de Zoologia, Museu Paraense Emilio Goeldi, Avenida Perimetral 1901, Terra
Firma, CEP 66077 530, Bel´
em, Par´
a, Brazil
3Universidad Tecnol´
ogica de Pereira, Facultad de Ciencias Ambientales, Vereda La Julita, AA 97, 660003 Pereira, Risaralda, Colombia
4COSTEL, UMR CNRS 5654, Universit´
e de Rennes 2, 5 place Henri Le Moal, 35 000 Rennes, France
5LEHNA, UMR 5023, Universit´
e Lyon 1, 43 Bd du 11 novembre 1918, 69 622 Villeurbanne Cedex, France
6Sorbonne Universit´
es, UPMC Univ Paris 06, iEES Paris, 4 place Jussieu, 75005, Paris, France
7SET, UMR CNRS 5603, Universit´
e de Pau et des Pays de l’Adour, Avenue du Doyen Poplawski, 64 000 Pau, France
8Insectos de Colombia, Yopal, Casanare, Colombia
9Embrapa Florestas, Estrada da Ribeira, km. 111, Caixa-Postal 319, CEP 83411 000, Curitiba, Parana, Brazil
10Universidad de la Amazonia, Sede Principal, Calle 17 – Diagonal 17 con Carrera 3F, Barrio Porvenir, Florencia, Caquet´
a, Colombia
11IRD, iEES Paris, Centre IRD Ile de France, 32 Av. Henri Varagnat, 93143 Bondy Cedex, France
12UR B&SEF, CIRAD, Campus international de Baillarguet, 34 398 Montpellier Cedex 5, France
13Agricultural Economics and Policy group, ETH Z¨
urich, Z¨
urich, Switzerland
14CIRAD, UPR Syst`
emes de p´
erennes, F-34398 Montpellier, France
15IRD, UMR Espace-DEV, MTD 500 rue Jean Franc¸ois Breton, 34 093 Montpellier, Cedex 5
16Universidade Federal Rural da Amazˆ
onia (UFRA), Avenida Perimetral 2501, CEP 66077 530, Bel´
em, Par´
a, Brazil
17Institut de Syst´
ematique, Evolution, Biodiversit´
e (ISYEB), Mus´
eum national d’Histoire naturelle, CNRS, Sorbonne Universit´
eEPHE,
57 rue Cuvier, CP 50, 75005 Paris, France
18Universidad Nacional de Colombia, Sede Palmira, Carrera 32 No 12 – 00 Chapinero, V´
ıa Candelaria, Palmira, Valle del Cauca,
Colombia
19Universidad del Valle, Cali, Colombia
20Instituto Tecnol´
ogico del Putumayo, Barrio Lu´
ıs Carlos Gal´
an, Mocoa, Putumayo, Colombia
21Centro Internacional de Agricultura Tropical (CIAT) – IRD, Cali, Valle del Cauca, Colombia
Abstract: Assessing how much management of agricultural landscapes, in addition to protected areas, can
offset biodiversity erosion in the tropics is a central issue for conservation that still requires cross-taxonomic
and landscape-scale studies. We measured the effects of Amazonia deforestation and subsequent land-use
intensification in 6 agricultural areas (landscape scale), where we sampled plants and 4 animal groups
(birds, earthworms, fruit flies, and moths). We assessed land-use intensification with a synthetic index based
on landscape metrics (total area and relative percentages of land uses, edge density, mean patch density and
diversity, and fractal structures at 5 dates from 1990 to 2007). Species richness decreased consistently as
agricultural intensification increased despite slight differences in the responses of sampled groups. Globally,
in moderately deforested landscapes species richness was relatively stable, and there was a clear threshold
in biodiversity loss midway along the intensification gradient, mainly linked to a drop in forest cover and
email: thibaud.decaens@cefe.cnrs.fr
Article impact statement: Tropical landscapes should retain over 40% of forests and 50% of unaffected patches to efficiently conserve biodiversity.
Paper submitted May 25, 2017; revised manuscript accepted May 16, 2018.
1380
Conservation Biology, Volume 32, No. 6, 1380–1391
C
2018 Society for Conservation Biology
DOI: 10.1111/cobi.13206
Deca¨
ens et al. 1381
quality. Our results suggest anthropogenic landscapes with high-quality forest covering >40 % of the surface
area may prevent biodiversity loss in Amazonia.
Keywords: biodiversity conservation, biodiversity erosion, Brazil, Colombia, landscape intensification, land-use
changes, threshold
P´
erdida de la Biodiversidad a lo largo de un Gradiente de Deforestaci´
on en Paisajes Agr´
ıcolas del Amazonas
Resumen: La evaluaci´
on sobre cu´
anto puede compensar el manejo de los paisajes agr´
ıcolas, adem´
as de las
´
areas protegidas, a la erosi´
on de la biodiversidad en los tr´
opicos es un tema central dentro de la conservaci´
on, el
cual todav´
ıa requiere estudios trans-taxon´
omicos y a escala de paisaje. Medimos los efectos de la deforestaci´
on
amaz´
onica y la intensificaci´
on subsecuente del uso del suelo en seis ´
areas agr´
ıcolas (escala de paisaje), en
donde muestreamos plantas y cuatro grupos faun´
ısticos (aves, lombrices de tierra, moscas de la fruta, y
polillas). Evaluamos la intensificaci´
on del uso de suelo con un ´
ındice sint´
etico basado en medidas de paisaje
(´
area total y porcentajes relativos del uso de suelo, densidad de borde, densidad y diversidad media de los
fragmentos, y estructuras fractales en cinco fechas desde 1990 hasta 2007). La riqueza de especies disminuy´
o
de forma constante conforme increment´
o la intensificaci´
on agr´
ıcola a pesar de las diferencias m´
ınimas en
las respuestas de los grupos muestreados. En general, la riqueza de especies en paisajes moderadamente
deforestados se mantuvo relativamente estable, y hubo un umbral distintivo en la p´
erdida de biodiversidad
a la mitad del gradiente de intensificaci´
on, principalmente vinculado con una ca´
ıda en la calidad y en la
cobertura forestal. Nuestros resultados sugieren que los paisajes antropog´
enicos con una cobertura forestal
de alta calidad >40% de la superficie pueden prevenir la p´
erdida de biodiversidad en el Amazonas.
Palabras Clave: brasil, cambios en el uso de suelo, Colombia, conservaci´
on de la biodiversidad, erosi´
on de la
biodiversidad, intensificaci´
on del paisaje, umbral
Introduction
Tropical rainforests support some of the greatest
diversities of living organisms on Earth (Myers 1984).
The Amazonian forest in particular has been popularized
as the last frontier forest and one of the most diverse
regions in the world (Wilson 2002; Cardoso Da Silva
et al. 2005). It is, however, dramatically threatened
by ongoing deforestation resulting from expanding
timber production and agricultural frontiers and by the
increased frequency of large-scale forest fires (Peres
et al. 2010; Newbold et al. 2015). Recent estimates
report a loss of about 770,000 km2of forest from 1970
to 2016, nearly 20% of the original forest cover (Butler
2017). This gradual destruction of forest ecosystems and
the resulting fragmentation and degradation of forest
remnants are strongly affecting Amazonian biodiversity
and the environmental sustainability of the whole region
(Haddad et al. 2015; Barlow et al. 2016).
The protection of pristine forests in integral protected
areas is vital to prevent loss of biodiversity. However,
their current coverage remains limited and their integrity
is frequently threatened in areas undergoing widespread
deforestation (Rodrigues et al. 2004). They are also not
representative of most of the tropical world, where the
vast majority of tropical rainforests are not comprised
within reserves and parks (Chazdon et al. 2009) and
>70% of remaining patches are within 1 km of the forest
edge (Haddad et al. 2015). It is therefore important to
focus not only on these protected areas, but also on the
conservation value of agricultural landscapes, which are
now replacing the pristine forest matrix at an increasing
pace (Barlow et al. 2007a; Chazdon et al. 2009; Gardner
et al. 2009).
Significant theoretical and empirical advances have
been made during the past decade in understanding of
the relative importance of different landscape proper-
ties in explaining biodiversity loss in tropical rainforest
biomes. For instance, recent studies emphasize the im-
portance of habitat amount (Melo et al. 2017), landscape
and within-forest disturbances (Barlow et al. 2016), and
landscape configuration (Villard & Metzger 2014) to ex-
plain the decline of biodiversity following deforestation.
However, most of these studies considered only a single
or a limited number of data-rich taxonomic groups, with
a strong bias toward vertebrates or plants (Fazey et al.
2005; Gardner et al. 2009). Understanding of the relative
sensitivity of different taxonomic groups to disturbance
at the landscape level remains insufficient (Dunn 2004);
some important components of terrestrial biodiversity,
such as invertebrates, are weakly documented (Collen
et al. 2009; Deca¨
ens 2010).
Although ecological theory predicts a global decrease
of biodiversity as ecosystem fragmentation increases
(Andren 1994; Wilson 2002), the pattern of this decrease
is still debated (Estavillo et al. 2013). Indeed, taxon- and
guild-specific responses to the degradation or transfor-
mation of tropical forest have been reported repeatedly
(Schulze et al. 2004; Barlow et al. 2007a;Newboldetal.
2014). Such idiosyncratic patterns may seriously hinder
the ability to assess the global impact of deforestation
when considering a limited number of indicator taxa.
Comparative studies conducted at the landscape scale
and across a wide range of taxa are thus critically
Conservation Biology
Volume 32, No. 6, 2018
1382 Biodiversity thresholds
needed to accurately assess the patterns of biodiversity
loss along deforestation and agricultural intensification
gradients and to identify the environmental drivers and
potential thresholds and tipping points in the response of
biodiversity to these perturbations (Newbold et al. 2014).
We measured the effects of Amazonia deforestation
and the subsequent land-use intensification on seven
different taxa including plants, vertebrates, and inver-
tebrates from above and below ground habitats at the
landscape scale. Our main hypothesis was there exists a
threshold in the response of biodiversity to gradual de-
forestation and landscape intensification beyond which
species are extirpated at a higher rate. We thus expected
biodiversity to decrease nonlinearly along a gradient of
deforestation and agricultural intensity. This assumption
was tested using a synthetic index of land-use intensity
and specific landscape metrics related to habitat loss,
habitat degradation, and land-use temporal dynamics.
Methods
Study Sites
The study was carried out in Brazil (state of Par´
a) and
Colombia (department of Caquet´
a) (Fig. 1). We studied
3 areas per country that represented 6 different stages
along a gradient of deforestation and agricultural intensifi-
cation, from recently affected areas, where deforestation
was in the early stage (Brazil), to regions with old agricul-
tural histories and extensive pastures (Colombia) (Fig. 1
& Supporting Information). We did not examine pristine
forests (i.e., the starting point of the deforestation pro-
cess) because the extensive agricultural landscapes be-
hind the Amazonian deforestation front no longer contain
large areas of intact forests. In a socioeconomic survey
X. Arnaud de Sartre categorized agricultural systems at
51 farms in each area (306 systems total) (see Supporting
Information). We then selected from this initial pool a set
of 54 farms (9 noncontiguous farms per area) that best
represented the local production systems (Lavelle et al.
2016a). We further used this set of 54 farms as replicated
elementary units to describe the landscape mosaic of the
entire landscape and to measure species richness for the
different target groups. At each farm, 5 equidistant sam-
pling points (200 m from each other) were geolocated
along a 1-km transect that ran along the longest diagonal
line of the farm plot (Fig. 1). This design guaranteed a
distribution of sampling effort among land uses that was
proportional to relative representation of the uses within
each farm.
Sampling
We used standardized methods to sample plants and 4
groups of animals (earthworms, fruit flies, moths, and
birds) on each farm. Groups were chosen for their link to
soil ecosystem services (earthworms, plants [Diaz et al.
2007; Lavelle et al. 2016b]), sensitivity to local pertur-
bations (fruit flies [Gottschalk et al. 2007]), or specific
response to landscape composition and migrating abili-
ties (moths and birds [Barlow et al. 2007b;Hawesetal.
2009]). Sampling was carried out from April to July 2008
with the aim of providing accurate and comparable esti-
mates of species richness among the 6 study areas.
At each sampling point, a quadrat sampling area of 50
×50 m was defined within a homogeneous landscape
element. For each taxonomic group, we determined
the species composition and counted the number of
individuals.
At each sampling point, plant species were counted in
3 vegetation strata in an area of a particular size: trees
>10 cm dbh in a 10 ×50 m area; shrubs and young
trees <10 cm diameter and >2mtallina5×50 m area;
herbaceous species (spermatophytes and pteridophytes)
and seedlings of both trees and shrubs in ten square me-
ters separated 5 m from each other. For each species,
samples were collected, dried in a field oven, and taken
to the lab for identification.
Saturniidae and Sphingidae (moths) were collected
with light trapping during a single collecting night at
each transect for a total of 9 collecting night per study
area. Traps were located as close as possible to the largest
forest remnant on the farm. Moths were sampled through-
out the night on moonless nights (i.e., from 1800 to 0600
[Lamarre et al. 2015]) with a white sheet (2 m height ×
3 m width) illuminated by a 175W-mercury vapor bulb
powered by a small portable generator. Specimens were
killed by injection of ammonia, stored and dried in la-
beled paper envelopes, and brought to the laboratory for
identification.
Drosophilidae (fruit flies) were collected with cylindri-
cal traps (10 cm diameter ×25 cm height) closed at one
extremity with a funnel at the other. Traps were baited
with approximately 100 g of fermented banana (Martins
et al. 2008). Three traps were placed at each sampling
point and deployed for 48 hours. Specimens were fixed
in 70% alcohol and taken to the laboratory for identifica-
tion. When necessary, specimens were kept alive to allow
laboratory rearing and more precise identifications.
Earthworms were collected by hand sorting 3 blocks
of soil at each sampling point: 1 central block of 25 ×
25 ×30 cm and 2 blocks of 25 ×25 ×10 cm (1 block
5 m north of the central block and the other 5 m south
of the central block). Individual specimens were fixed in
4% formaldehyde and taken to the laboratory for identifi-
cation. Because most of the collected species were new
to science, morphospecies were defined on the basis of
external and internal morphological characters, and the
number of morphospecies was used as a surrogate for
species richness.
Conservation Biology
Volume 32, No. 6, 2018
Deca¨
ens et al. 1383
(a)
(c)
(d)
(b)
Figure 1. Study sites in Amazonia and sampling design: (a) location of the 6 sites in Brazil and Colombia selected
to represent a gradient of deforestation and land-use intensity, (b) shape of each study area; (c) examples of farms
selected in a given study area (here Mac¸aranduba) and locations of the 5 sampling points for biodiversity
assessment; (d) example of how the landscape structure on 1 farm was described at 5 different dates to account
for landscape dynamics.
Conservation Biology
Volume 32, No. 6, 2018
1384 Biodiversity thresholds
Point counts for birds were conducted at each sam-
pling point by two persons in a single 20-minute session
in the morning (from approximately 0600 to 1100) to
encompass the period of maximal bird activity. The total
sampling effort was of 45 bird counts per study area.
Relying on a single sampling period and a single season
could have led to undersampling of rare species or taxa
with strong seasonal occupancy patterns. However, we
considered our sampling strategy relevant to efficiently
sampling, for each focal group, at least the dominant
species that are active during the rainy season. This al-
lowed us to conduct a broad comparison of diversity
levels among the different study areas.
Species Richness
Regional species richness was described by calculating
the cumulative number of species observed in each
area and the abundance-based coverage estimator (ACE),
which provides an estimate of the theoretical size of the
species pool (Chao & Lee 1992). For rarefaction and ex-
trapolation curves, we plotted the number of species ac-
cording to sampling intensity (i.e., number of specimens
collected or observed) and extrapolated this number in a
hypothetical situation where the sampling size was dou-
bled. This was done using the iNEXT package for R (R
Development Core Team 2011; Chao & Jost 2012).
At the farm scale, local species richness was rarefied
for a subsample size that corresponded to the median
density of each group of organism. For each farm, species
richness transformed in this way represented the average
number of species observed in a standardized sample
of nindividuals, where nis the median density of the
corresponding target group and nequaled 1586, 219, 35,
107, 30, 29, and 140 for herbaceous plants, shrubs, trees,
fruit flies, moths, earthworms, and birds, respectively.
Calculations were done using the vegan library of the R
software (Oksanen et al. 2008).
At both scales, we computed a standardized index
of species richness, modified from Dunn (2004), that
reduced most of the species richness variation into a
single metric. We transformed each value of estimated
regional richness (ACE index) or local rarefied richness
as a percentage of the highest value observed at the rel-
evant scale for the corresponding taxonomic group. The
resulting data set was composed of normalized species
richness that ranged from 0 to 100 for each group of
organisms. The standardized index of species richness
was calculated for each farm and each area as the mean
of the transformed richness of the seven focal groups;
it also ranged from 0 (theoretical farm or area where
species richness was null for all the groups) to 100 (the-
oretical farm or area where species richness in all groups
were the highest observed values). Pairwise correlations
in species richness among the focal groups of organisms
were assessed with Spearman correlation index.
Landscape Effects on Species Richness
We used remote sensing to describe landscape compo-
sition and structure within each farm. Our set of met-
rics included total area of land use and relative percent-
ages, edge density, mean patch density, patch diversity
(patch richness, Shannon diversity, Shannon evenness,
and dominance), and fractal structures (perimeter/area,
mean shape). We used available Landsat images for 1990,
1994, 1998, 2002, and 2007. These metrics were fur-
ther used to calculate a single index of land-use inten-
sity (LI) (Oszwald et al. 2011) that integrates the current
pattern and the dynamics of the landscapes (Supporting
Information).
We also used these metrics to assess the specific effect
of five potential landscape drivers (Table 1) on species
richness: forest loss based on relative cover of mature
forests (percentage of farm area); agricultural intensifi-
cation based on the relative cover of agricultural land
(percentage of farm area covered by pastures and crops);
landscape fragmentation based on total edge density,
which appeared to be significantly correlated with other
metrics describing landscape structure and diversity
(Oszwald et al. 2011); forest quality based on the pro-
portion of forest in a given farm that was undisturbed
during the past 17 years (hereafter undisturbed forest);
and landscape dynamics, calculated using Euclidean dis-
tance with the daisy function in the R cluster library as
the mean between-year land-use turnover (i.e., change
in land-use composition) from 1990 to 2007 (Maechler
et al. 2012). See Oszwald et al. (2011) and Supporting
Information for further details.
Co-inertia analysis was conducted to assess the relation-
ships between the rarefied richness of the 7 focal groups
of organisms and the landscape drivers measured in the
54 farms. Co-inertia analysis consists of a simultaneous
ordination of 2 tables (for instance an environmental and
a floristic/faunistic table) that share the same row codes
and provides axes that maximize the covariation between
them (Dol´
edec & Chessel 1994; Dray et al. 2003). In the
present case, we used the ADE-4 package for R (Dray
& Dufour 2007) to independently performed principal
component analysis (PCA) on each table and used their
results to run the co-inertia analysis. This allowed a si-
multaneous ordination of the 7 focal groups according
to the 5 landscape drivers in the 54 farms and provided
a measure of the global similarity among the data sets
(matrix correlation coefficient Rv) that we tested using a
Monte Carlo permutation test (999 randomizations).
Graphical exploration of the data suggested the rela-
tionship among local species richness, land-use inten-
sity, and individual landscape metrics was often nonlin-
ear and was characterized in some cases by an abrupt
change along predictors. Thus, we adjusted species rich-
ness to landscape metrics with 3 kinds of models and
selected the one with the best AIC. The 3 models
Conservation Biology
Volume 32, No. 6, 2018
Deca¨
ens et al. 1385
Table 1. Main characteristics of the Amazonian landscapes in the 6 agricultural areas examined.
Land cover 2007 (%)bMetricsc
Landscape
(code)
Colonization
(years)aFO BF UW PA LP FA WA BS Fqual (%) ED (m/ha) LUTO (%) LI
Pacaj´
a (A) 10 77.1 19.4 1.6 0.2 0.1 1.5 0.0 0.1 0.88 476.2 19.8 0.21
Palmares (B) 20 81.7 9.4 1.4 0.1 0.0 7.0 0.0 0.3 0.78 689.0 15.0 0.38
Mac¸aranduba (C) 40 22.7 10.9 15.6 10.0 12.8 27.9 0.0 0.0 0.49 622.6 31.9 0.51
Aguadulce (D) >80 1.5 0.0 45.1 7.9 10.1 14.4 20.9 0.0 0.06 738.3 28.6 0.74
Balcanes (E) >80 4.2 0.0 18.7 11.8 29.7 5.1 30.3 0.1 0.17 698.0 34.7 0.85
Canelos (F) >80 2.0 0.0 19.1 18.2 25.4 11.8 23.3 0.2 0.09 653.3 31.0 0.80
aAge of human colonization.
bLand-cover percent calculated from 2007 Landsat images: FO, primary forests; BF, burned forests; UW, undergrowth wetlands; PA, pastures; LP,
ligneous pastures; FA, fallows; WA, water areas; BS, bare soil.
cMeasures of landscape structure: Fqual, forest quality (percent of forest in a given area that was undisturbed during the past 17 years), ED,
total edge density; LUTO, land-use turnover (i.e., mean between-year dissimilarity in land-use composition from 1990 to 2007); LI, land-use
index (0, lowest intensity; 1, highest intensity).
included simple linear effect of a predictor, segmented
linear model effect, or nonlinear logistic model effect.
Segmented analyses automatically estimating the break
point in the predictor were performed with the seg-
mented and MASS package of R. Linear and logistic re-
gressions were done using the lm and SSlogis functions of
the stats package. Only the best model for each predictor
is shown on figures, and nonsignificant models are not
shown.
Results
In total, we sampled or recorded >3,800 species, includ-
ing 100 species of fruit-flies, 136 species of moths, 338
species of birds, 21 species of earthworms, and 1746,
1049, and 414 species of herbaceous plants, shrubs,
and trees, respectively. At the regional scale (Fig. 2 &
Supporting Information), the general pattern was a tran-
sient stability of species pools at the beginning of the
landscape gradient (i.e., areas of less intensive agricul-
ture). In moderately degraded areas there was sometimes
a slight increase in the number of species (e.g., fruit flies
and earthworms) (Figs. 2d and 2f), followed by a sharp
decrease beyond a threshold leading to poorly diversi-
fied assemblages in landscapes with the highest land-use
intensity. The exact position of the threshold on the gra-
dient was different among the target groups, occurring,
for example, at low levels of land-use intensity for herba-
ceous plants and moths (Fig. 2a and e) and at the most
degraded extreme of the gradient for trees (Fig. 2c). The
synthetic index of regional richness efficiently summa-
rized these patterns and highlighted a clear threshold in
the loss of regional richness mid-way along the land-use
intensity gradient (Fig. 2h).
With the exclusion of birds, we found a general de-
crease in local richness along the landscape gradient
for all our target groups (Fig. 3). The rarefied richness
of herbaceous plants and earthworms decreased rather
monotonically (Figs. 3a and c). Shrubs, trees, fruit flies,
and moths presented a transient stability in their rich-
ness and a marked drop in richness at moderate land-use
intensity (Figs. 3b–e). As for regional diversity, local rich-
ness appeared rather constant in the first stages of forest
conversion before a clear threshold was reached, mid-
way along the gradient, beyond which diversity declined
sharply (Fig. 3h).
Co-inertia analysis highlighted a significant costructure
between species richness and landscape metrics (Rv co-
efficient =0.44; Monte Carlo simulated p=0.001).
The first co-inertia axis explained 99.6 % of the covari-
ation among data sets. It separated farms with high for-
est cover and quality (mostly Brazilian sites) from farms
with a high percentage of agricultural land cover and
to a lesser extent rapid land-use turnover and high edge
density (negative scores, mostly Colombian sites) (Figs.
4a and SI4). Rarefied species richness of all groups of
organisms had positive scores along this axis (Fig. 4b),
meaning biodiversity was globally and positively linked
with the availability and quality of forests in the landscape
and negatively linked with agriculture intensification.
Accordingly, the correlation between local species
richness and the percentage of forest ecosystems showed
richness stability in moderately deforested farms (40–
50% of forest dominance) and a clear threshold between
30% and 40% of forest coverage below which biodiversity
tended to decline rapidly (Fig. 5a). Differences between
farms with comparable forest cover were further ex-
plained by differences in the quality of the forest patches,
here defined as the percentage of undisturbed forests
(Fig. 5b). We also observed an almost linear decrease
of local richness as the proportion of agricultural lands
increased in the landscape (Fig. 5c). No significant rela-
tionship occurred between local richness and land-use
turnover or total edge density (Supporting Information).
Although positive and significant correlations of local
species richness occurred with all possible pairwise
comparisons between groups (Supporting Information),
Conservation Biology
Volume 32, No. 6, 2018
1386 Biodiversity thresholds
ABCDEF
(a) Herbaceous plants
0
200
400
600
800
1000
No. of species
ABCDEF
(d) Fruit flies
0
10
20
30
40
No. of species
ABCDEF
(g) Birds
0
50
100
150
200
No. of species
ABCDEF
(b) Shrubs
0
100
200
300
400
500
ABCDEF
(e) Moths
0
20
40
60
80
100
120
ABCDEF
(c) Tree s
0
50
100
150
200
ABCDEF
(f) Earthworms
0
5
10
15
ABCDEF
(h) All groups
0
20
40
60
80
100
Richness index (%)
Figure 2. Variations in regional species richness with land-use intensity for: (a–g) 7 target groups and (h) overall
regional biodiversity calculated for all groups with the standardized species richness index (6 landscape areas
ranked by increasing level of land-use intensity (see Table 1) (A, Pacaj´
a; B, Plamares; C, Mac¸aranduba; D,
Aguadulce; E, Balcanes; F, Canelos; hash-marked bars, cumulative number of species observed in each area; bars
without hash marks, estimated regional richness calculated with the abundance-based coverage estimator;
vertical lines, SE).
the average correlation was low (mean r=0.54). Birds
and moths were poorly correlated with other groups
(0.33 and 0.41, respectively), whereas other groups of
animals had higher correlations (0.57 and 0.58 for fruit
flies and earthworms, respectively). Plant richness was
more closely correlated to the other groups (0.61 for
the herbaceous layer and 0.63 for shrubs and trees).
Average between-group correlations of ACE-estimated
regional richness was higher than that observed at the
local scale (mean r=0.71), but pairwise comparisons
were significant in only 7 cases out of 21 (Supporting
Information). For the comparison between shrub and
herbaceous plant layers, pwas <0.01, whereas other sig-
nificant pairwise comparisons had pvalues from 0.01 to
0.05.
Discussion
We identified a dramatic impact of land-use intensity on
Amazonian biodiversity. The clear threshold we found
in the response of cross-taxonomic species richness to
land-use intensity represents significant empirical sup-
port for the existence of tipping points in the response
of biodiversity to deforestation and land-use intensity.
Although Brazilian and Colombian sites may not have
originally equivalent species diversity, this threshold is
supported for some focal taxa within species pools of
equivalent size in both regions, or even larger species
pools in Colombia (Supporting Information). This find-
ing suggests loss of species richness could have been
underestimated in the higher range of our gradient. Such
Conservation Biology
Volume 32, No. 6, 2018
Deca¨
ens et al. 1387
0.2 0.4 0.6 0.8
0
50
100
150
200
(a) Herbaceous plants
No. of species
0.2 0.4 0.6 0.8
0
5
10
15
20
(d) Fruit flies
No. of species
0.2 0.4 0.6 0.8
0
10
20
30
40
50
60
(g) Birds
No. of species
0.2 0.4 0.6 0.8
0
20
40
60
80
100
(b) Shrubs
0.2 0.4 0.6 0.8
0
5
10
15
20
25
(e) Moths
0.2 0.4 0.6 0.8
0
5
10
15
20
25
30
(c) Trees
0.2 0.4 0.6 0.8
0
2
4
6
8
10
(f) Earthworms
0.2 0.4 0.6 0.8
0
20
40
60
80
100
(h) All groups
Land use intensity
Richness index (%)
A
B
C
D
E
F
Figure 3. Relationships between local species richness and land-use intensity (0, lowest intensity; 1, highest
intensity): (a–g) rarefied richness calculated for 7 target groups and (h) overall richness calculated for all groups
with the standardized richness index (points, average values obtained from the different farms in each landscape).
a nonlinear decrease in species richness following defor-
estation is predicted by ecological theory (Andren 1994;
Wilson 2002; Lindenmayer et al. 2008), but little empiri-
cal evidence has been collected so far to test for such a
pattern, especially at the landscape scale. To date, the va-
lidity of the threshold hypothesis has been questioned by
authors who believe minor changes in native vegetation
cover will not result in abrupt biodiversity changes due to
contrasting responses of individual species (Lindenmayer
et al. 2008; Pardini et al. 2010). However, in spite of the
variety of the groups considered, we found that over-
all local biodiversity dropped steeply when forest cover
fell below 30–40%. This threshold value is in the higher
range of those predicted by Andren (1994), who hypoth-
esized the existence of a universal threshold value around
10–30% of native vegetation cover. It is also consistent
with the results of Rigueira et al. (2013), who found a
6-fold decrease in Myrtaceae species richness when for-
est cover drops below 30–40% in the Brazilian Atlantic
Forest. We also found that forest patch quality was an
important factor in explaining biodiversity dynamics in
deforested landscapes, which is consistent with the re-
sults of Barlow et al. (2016), who highlighted that forest
disturbance in individual forests and at the landscape
level can explain a significant proportion of conservation
value loss in Amazonian landscapes.
Landscape structure (total edge density) and temporal
dynamics did not explain biodiversity patterns, which
suggests habitat amount may have greater effects
than habitat fragmentation (Andren 1994). This is an
unexpected result because the link between landscape
composition and fragmentation is well established
especially in the context of deforestation fronts, where in-
tensification of land use results in an increase in anthropic
ecosystems and a reduction of the average size of land-
scape units (and consequently an increase in total edge
Conservation Biology
Volume 32, No. 6, 2018
1388 Biodiversity thresholds
(a) Landscape metrics
Forests
Agroecos.
Turn over
Edge dens.
Forest qual.
(a) Landscape metrics (b) Rarefied species richness
Fruit flies
Moths
Birds
Earthworms
Herbaceous
Shrubs
Trees
(b) Rarefied species richness
Figure 4. Contributions of (a) landscape metrics and (b) rarefied species richness to the definition of the first 2
axes of the co-inertia analysis (qual, quality).
density). Our finding that temporal dynamics did not ex-
plain biodiversity patterns also contrasts with the results
of other studies that show patch temporal turnover does
affect biodiversity (Fischer 2001; Ernoult et al. 2006). One
explanation for this is that we considered several taxa
with potentially different ecological responses to land-
scape dynamics, which could have blurred the global re-
sponse of biodiversity to landscape structure and land use
turnover.
Besides the clear threshold for loss of overall species
richness as land use intensifies, significant variations in
the response patterns among our focal groups were ap-
parent. For instance, the species richness of herbaceous
plants, trees, and earthworms dropped sooner at the local
than at the regional scale, suggesting that forest patches
allowed these groups to maintain transient pools of
species regionally or that the extinction of forest species
was temporarily compensated by the invasion of extrare-
gional species adapted to agricultural land uses. By con-
trast, shrubs, fruit flies, and birds presented simultaneous
variations of their regional and local richness with land-
use intensity, indicating that local species extirpations
rapidly trigger extirpations at the regional scale. Moths
showed a rather idiosyncratic pattern; regional richness
decreased more rapidly than local richness, which may be
explained by a transient decrease in community turnover
at moderate levels of land-use intensity, allowing for a
higher proportion of the regional species pool to occur
in local assemblages.
Despite positive correlations among species richness
of our different target groups, our results highlighted that
these correlations were mostly nonsignificant at the re-
gional scale and that only a small fraction of the variance
in local species richness of one taxonomic group enabled
prediction of species richness of another group. These
observations follow the conclusions of previous studies
conducted in other tropical environments (Lawton et al.
1998; Schulze et al. 2004). Discrepancies in the response
of different taxa to deforestation can be caused by dif-
ferences in their sensitivity to landscape transformation
because they experience landscape changes at different
spatial and temporal scales, are affected differently by
changes in landscape configuration, use different habitats
and resources within these landscapes, and present dif-
ferent dispersal capabilities to offset the effects of habitat
fragmentation (Montoya et al. 2008; Gardner et al. 2009;
Villard & Metzger 2014). Therefore, species extinctions
tend to occur at different levels of landscape perturba-
tion, and the general response of species richness may
differ significantly among the target groups. Altogether,
these results emphasize the importance of considering a
large range of organisms, rather than focusing on a few
indicator taxa, when addressing the question of biodi-
versity response to forest modification and destruction
(Lawton et al. 1998; Barlow et al. 2007a). The use of a
synthetic index of species richness represents an efficient
tool to summarize both the constancy and variability of
the patterns among groups in a single and easily manage-
able metric.
A comparison of the total number of species observed
in our samples with the estimated sizes of regional species
pools for the few taxonomic groups for which this infor-
mation is available suggests that an important proportion
of the species potentially occurring in the surveyed re-
gions was lacking in our study (Supporting Information).
In the less affected landscapes, however, observed and
Conservation Biology
Volume 32, No. 6, 2018
Deca¨
ens et al. 1389
0 20 40 60 80 100
0
20
40
60
80
100
Forest (%)
Richness index (%)
(a)
0.0 0.2 0.4 0.6 0.8 1.0
0
20
40
60
80
100
Forest quality
(b)
0 20 40 60 80 100
0
20
40
60
80
100
Agricultural land (%)
Richness index (%)
(c)
A
B
C
D
E
F
Figure 5. Relationships between the synthetic index of species richness and selected landscape metrics and (a)
percent forest area, (b) forest quality, and (c) percent agricultural lands.
estimated numbers of species for trees, fruit flies, moths,
and birds were higher or in the range of those numbers
found locally by Barlow et al. (2007a) in primary Ama-
zonian forest based on a comparable sampling intensity.
This suggests that although less intense agricultural land-
scapes are likely to sustain high levels of species richness,
a number of forest species are lacking in those areas prob-
ably as a consequence of local extirpations occurring very
early during the onset of the deforestation process. In
agricultural landscapes, our results, however, underline
the importance of old secondary vegetation, managed
forests, and tree plantations in the maintenance of local
species richness for different groups of plants and animals
(Lawton et al. 1998; Barlow et al. 2007a).
Finally, our findings improve substantially understand-
ing of the importance of landscape structure and dynam-
ics for the maintenance of biodiversity in human-affected
areas. The sharp threshold observed for the biodiversity
response to land-use intensity is mainly explained by for-
est availability and quality. Species richness was not main-
tained when forest cover fell below 40% and when forest
patches contained <50% of undisturbed forest. An inter-
esting correlate is that biodiversity loss may be mitigated
either by sustaining forest quality (maintaining a high
proportion of primary forest in the remaining patches)
in highly deforested landscapes or by maintaining large
forest surfaces in regions where forest has been exten-
sively degraded. Landscape fragmentation and dynamics
seemed of lesser importance for the taxa we studied. Our
findings therefore suggest that anthropogenic landscapes
with high-quality forests covering >40%ofthesurface
area (Figs. 5a & b) may have a significant potential to
offset biodiversity loss in Amazonia.
Although this result provides valuable clues for iden-
tifying landscape management options toward meeting
the challenges of biodiversity conservation and economic
development in rural areas, it could be mistakenly used
to justify the development of agriculture in still intact
Conservation Biology
Volume 32, No. 6, 2018
1390 Biodiversity thresholds
forest areas. The applicability of our findings must be
considered with caution and should not be considered
outside the context of already degraded landscapes be-
cause the maintenance of species richness alone may not
represent a sufficient indicator of conservation success
(Barlow et al. 2007a). For instance, the use of biodiver-
sity metrics to measure the dissimilarity in community
composition between intact forests and anthropogenic
landscapes or to examine species life-trait shifts along de-
forestation gradients, although not possible in our study,
should be considered in future cross-taxonomic studies. If
we could have conducted such investigations, we would
have been able to determine the relative importance of
intact forest to species along the gradient of agricultural
development or to determine species that are particularly
sensitive to forest degradation. Additionally, the presence
of a threshold at intermediate levels of deforestation does
not exclude the possibility of an early drop in biodiversity
at the onset of the deforestation process; we did not
considered this.
However, identifying relevant landscape characteris-
tics for biodiversity conservation in agricultural lands
paves the way to a definition of management policies for
the reconstruction of sustainable landscapes, which will
help reduce biological erosion by complementing the
existing networks of protected forest areas. For instance,
both Colombia and Brazil have the political will to pre-
serve their forest resources and reforest already degraded
areas (e.g., 20 ×20 initiative in Colombia and the revised
Forest Code in Brazil [Casa Civil 2012]). However, exist-
ing policies are often not applicable (Landau et al. 2012;
Vedovato et al. 2016) and do not consider the quality of
forest patches (e.g., up to 50% of non-native tree species
are allowed in reforested areas in Brazil). Our results sug-
gest that current legislation imposes fragile conditions
on the conservation of biodiversity in rural landscapes
and that the quality of forest remnants should be better
considered when dealing with biodiversity monitoring.
Acknowledgments
This work is dedicated to the late J. C. Ribeiro da Silva and
M. do Espirito Santo Da Silva whose farm was one of the
study sites of this work. Their murder on 24 May 2011
in Mac¸aranduba is doubtlessly tied to their strong en-
gagement in the preservation of biodiversity in Brazilian
Amazonia and the development of sustainable agriculture
in the region. We thank all the colleagues from the Federal
University of Par´
a (Maraba campus, Brazil), Universidad
de la Amazonia (Florencia, Colombia), Museu Paraense
Emilio Goeldi (Bel´
em, Brazil), and the Institut Recherche
pour le D´
eveloppement (Paris, France) for their critical
logistical support of this study. We are also grateful to
the Brazilian, Colombian, and French students who en-
thusiastically participated in field sampling and helped
process the enormous number of samples and data sets,
the Brazilian and Colombian technical and field staff, es-
pecially D. da Costa Carvalho and M. Cordeiro, and the
Brazilian and Colombian farmers for allowing us to carry
out our study. This study was supported through grants
from the French Agence Nationale de la Recherche and
the Brazilian National Council of Research to the project
AMAZ.
Supporting Information
Details on the selection of study sites and calculation of
the land use index (Appendix S1), questionnaire used
for farmers interviews in the socioeconomic survey of
agricultural systems (in Portuguese) (Appendix S2), and
species’ pool estimated sizes for some taxa, rarefaction
curves, additional results of the co-inertia analysis, non-
significant correlations between species richness and
landscape metrics, and pairwise correlations (Spearman’s
rank correlation) of local and regional species richness
among all groups of organisms (Appendix S3) are avail-
able online.
Literature Cited
Andren H. 1994. Effects of habitat fragmentation on birds and mam-
mals in landscapes with different proportions of suitable habitat - a
review. Oikos 71:355–366.
Barlow J, et al. 2007a. Quantifying the biodiversity value of tropical
primary, secondary, and plantation forests. PNAS 104:18555–18560.
Barlow J, Mestre LAM, Gardner TA, Peres CA. 2007b.Thevalueof
primary, secondary and plantation forests for Amazonian birds. Bio-
logical Conservation 136:212–231.
Barlow J, et al. 2016. Anthropogenic disturbance in tropical forests can
double biodiversity loss from deforestation. Nature 535:144–147.
Butler R. 2017. Calculating deforestation figures for the Ama-
zon. https://rainforests.mongabay.com/amazon/deforestation_
calculations.html.
Cardoso Da Silva JM, Rylands AB, Da Fonseca GAB. 2005. The fate of the
Amazonian areas of endemism. Conservation Biology 19:689–694.
Casa Civil, Presidˆ
encia da Rep´
ublica. 2012. Lei N°12.651, de
25 de maio de 2012. http://www.planalto.gov.br/ccivil_03/_ato
2011-2014/2012/lei/L12651compilado.htm.
Chao A, Jost L. 2012. Coverage-based rarefaction and extrapolation:
standardizing samples by completeness rather than size. Ecology
93:2533–2547.
Chao A, Lee S-M. 1992. Estimating the number of classes via sample
coverage. Journal of the American Statistical Association 87:210–
217.
Chazdon RL, et al. 2009. Beyond reserves: A research agenda for
conserving biodiversity in human – modified tropical landscapes.
Biotropica 41:142–153.
Collen B, Loh J, Whitmee S, McRae L, Amin R, Baillie JEM. 2009. Mon-
itoring change in vertebrate abundance: the Living Planet Index.
Conservation Biology 23:317–327.
Deca¨
ens T. 2010. Macroecological patterns in soil communities. Global
Ecology And Biogeography 19:287–302.
DiazS,LavorelS,deBelloF,Qu
´
etier F, Grigulis K, Robson TM.
2007. Incorporating plant functional diversity effects in ecosystem
Conservation Biology
Volume 32, No. 6, 2018
Deca¨
ens et al. 1391
service assessments. Proceedings of the National Academy of Sci-
ences 104:20684–20689.
Dol´
edec S, Chessel D. 1994. Co-inertia analysis: an alternative method
for studying species–environment relationships. Freshwater Biology
31:277–294.
Dray S, Chessel D, Thioulouse J. 2003. Co-inertia analysis and the linking
of the ecological data tables. Ecology 84:3078–3089.
Dray S, Dufour AB. 2007. The ade4 package: implementing the duality
diagram for ecologists. Journal of Statistical Software 22:1–20.
Dunn RR. 2004. Managing the tropical landscape: a comparison of the
effects of logging and forest conversion to agriculture on ants, birds,
and lepidoptera. Forest Ecology and Management 191:215–224.
Ernoult A, Tremauville Y, Cellier D, Margerie P, Langlois E, Alard D.
2006. Potential landscape drivers of biodiversity components in a
flood plain: Past or present patterns? Biological Conservation 127:1–
17.
Estavillo C, Pardini R, da Rocha PLB. 2013. Forest loss and the bio-
diversity threshold: an evaluation considering species habitat re-
quirements and the use of matrix habitats. PLOS ONE 8:e82369.
https://doi.org/10.1371/journal.pone.0082369.
Fazey I, Fischer J, Lindenmayer DB. 2005. What do conservation biolo-
gists publish? Biological Conservation 124:63–73.
Fischer M. 2001. Landscape dynamics can accelerate metapopulation
extinction. Trends in Ecology & Evolution 16:225–226.
Gardner TA, Barlow J, Chazdon RL, Ewers RM, Harvey CA, Peres CA,
Sodhi NS. 2009. Prospects for tropical forest biodiversity in a human-
modified world. Ecology Letters 12:561–582.
Gottschalk MS, De Toni DC, Valente VLS, Hofmann PRP. 2007. Changes
in Brazilian Drosophilidae (Diptera) assemblages across an urbani-
sation gradient. Neotropical Entomology 36:848–862.
Haddad NM, et al. 2015. Habitat fragmentation and its lasting impact on
Earth’s ecosystems. Science Advances 1(2):e1500052.
Hawes J, da Silva Motta C, Overal WL, Barlow J, Gardner TA, Peres
CA. 2009. Diversity and composition of Amazonian moths in pri-
mary, secondary and plantation forests. Journal of Tropical Ecology
25:281–300.
Lamarre G, Mendoza I, Rougerie R, Deca¨
ens T, H´
erault B, B´
en´
eluz F.
2015. Stay out (almost) all night: contrasting responses in flight
activity among tropical moth assemblages. Neotropical Entomology
44:109–115.
Landau EC, da Cruz RK, Hirsch A, Pimenta FM, Pereira Guimar˜
aes D.
2012. Variac¸˜
ao geogr´
afica do tamanho dos M´
odulos Fiscais no Brasil.
Documentos No. 146, Embrapa Milho e Sorgo, Sete Lagoas.
Lavelle P, et al. 2016a. Unsustainable landscapes of deforested Ama-
zonia: An analysis of the relationships among landscapes and the
social, economic and environmental profiles of farms at different
ages following deforestation. Global Environmental Change 40:137–
155.
Lavelle P, Spain A, Blouin M, Brown G, Deca¨
ens T, Grimaldi M, Jim´
enez
JJ, McKey D, Mathieu J, Velasquez E, Zangerl´
e A. 2016b. Ecosystem
engineers in a self-organized soil: a review of concepts and future
research questions. Soil Science 181:91–109.
Lawton JH, et al. 1998. Biodiversity inventories, indicator taxa and ef-
fects of habitat modification in tropical forest. Nature 391:72–76.
Lindenmayer D, et al. 2008. A checklist for ecological management of
landscapes for conservation. Ecology Letters 11:78–91.
Maechler M, Rousseeuw P, Struyf A, Hubert M, Hornik K. 2012. Cluster:
cluster analysis basics and extensions. R package version 1.14.3.
Martins MB, Pena JAN, Bittencourt RN. 2008. Traps for tropical
drosophilids survey. Drosophila Information Service 91:91–91.
Melo GL, Sponchiado J, C´
aceres NC, Fahrig L. 2017. Testing the habitat
amount hypothesis for South American small mammals. Biological
Conservation 209:304–314.
Montoya D, Zavala MA, Rodr´
ıguez MA, Purves DW. 2008. Animal versus
wind dispersal and the robustness of tree species to deforestation.
Science 320:1502–1504.
Myers N. 1984. The primary source: tropical forests and our future.
W.W. Norton, New York.
Newbold T, et al. 2014. A global model of the response of tropical and
sub-tropical forest biodiversity to anthropogenic pressures. Proc. R.
Soc. B Biol. Sci. 281:20141371.
Newbold T, et al. 2015. Global effects of land use on local terrestrial
biodiversity. Nature 520:45–50.
Oksanen J, Kindt R, Legendre P, O’Hara B, Simpson GL, Stevens MHH,
Wagner H. 2008. Vegan: Community Ecology Package. R package
version 1.13-11. http://vegan.r-forge.r-project.org/.
Oszwald J, Gond V, Dol´
edec S, Lavelle P. 2011. Identification
d’indicateurs de changement d’occupation du sol pour le suivi
des mosa¨
ıques paysag`
eres. Bois et Forˆ
ets des Tropiques 307:
7–21.
Pardini R, Bueno AD, Gardner TA, Prado PI, Metzger JP. 2010. Be-
yond the fragmentation threshold hypothesis: regime shifts in
biodiversity across fragmented landscapes. PLOS ONE 5:e13666.
https://doi.org/10.1371/journal.pone.0013666.
Peres CA, Gardner TA, Barlow J, Zuanon J, Michalski F, Lees AC, Vieira
ICG, Moreira FMS, Feeley KJ. 2010. Biodiversity conservation in
human-modified Amazonian forest landscapes. Biological Conserva-
tion 143:2314–2327.
R Development Core Team. 2011. R: A language and environment
for statistical computing. R Foundation for Statistical Computing,
Vienna, Austria. http://www.R-project.org.
Rigueira DMG, da Rocha PLB, Mariano-Neto E. 2013. Forest cover, ex-
tinction thresholds and time lags in woody plants (Myrtaceae) in
the Brazilian Atlantic Forest: resources for conservation. Biodiversity
Conservation 22:3141–3163.
Rodrigues ASL, et al. 2004. Effectiveness of the global protected
area network in representing species diversity. Nature 428:
640–643.
Schulze CH, et al. 2004. Biodiversity indicator groups of tropical land-
use systems: comparing plants, birds, and insects. Ecological Appli-
cations 14:1321–1333.
Vedavato LB, Fonseca MG, AraiE, Oighenstein Anderson L, Arag˜
ao
LEOC. 2016. The extent of 2014 forest fragmentation in the Brazilian
Amazon. Regional Environmental Change 16:2485–2490.
Villard M-A, Metzger J-P. 2014. Beyond the fragmentation debate: a con-
ceptual model to predict when habitat configuration really matters.
Journal of Applied Ecology 51:309–318.
Wilson EO. 2002. The future of life. Vintage Books, New York.
Conservation Biology
Volume 32, No. 6, 2018
... Nevertheless, the observed species richness in these three major regions, Orinoquia, Amazon, and Caribbean-which occupy more than 70 % of Colombia's continental surface-remains relatively low, most likely due to the lack of surveys in these areas since most studies have been focused on the Andes and Pacific regions. This suggests greatly underestimated Saturniidae diversity, especially in representative and critically endangered Colombian ecosystems, such as the Amazon rainforest and the tropical dry forests (TDF) of the national territory, both of which are significantly impacted by severe deforestation and habitat loss/degradation through multiple anthropogenic factors (García et al. 2014;Decaëns et al. 2018). ...
... These are Copaxa apollinairei Lemaire, 1978 from the Andes and Syssphinx chocoensis Lemaire, 1988 from Chocó department, both of which have been given VU conservation status due to the threat of habitat degradation and loss (Amarillo 2007). Such information is essential to propose effective conservation areas, a crucial issue for Colombian biodiversity, because many localities are rapidly disappearing as a result of indiscriminate deforestation and environmental degradation (García et al. 2014;Decaëns et al. 2018). ...
Article
Based on the review of literature and biological collections, information retrieved from public online databases, and from fieldwork conducted between 2015 and 2020, we provide more than 3500 occurrence records and an updated checklist of Colombian Saturniidae, annotated with distribution data for all species. In the first checklist of Colombian saturniids published two decades ago, a total of 184 species were cited; in the current update the number has risen to 653 species/subspecies classified in 55 genera, representing all six recognized Neotropical subfamilies. The Andean and Pacific regions are the richest, accounting for about three quarters of all species listed herein. We call attention to the fact that for most speciose saturniid genera in the Neotropics, there remains a significant need for further taxonomic and systematic research with objective and integrative approaches. Likewise, it is necessary to continue the study of Saturniidae moths in Colombia, strategically increasing the sampling efforts in specific under-sampled natural regions to further document the diversity of this family in the country.
... In addition to the above, studies focused on a single taxon may provide incomplete and less useful information for conservation plans, since the species differ in their sensitivity to habitat modification and so there are different responses in life cycle traits, such as dispersal capacity, reproductive potential and niche width (Díaz-García et al., 2020;Kellner et al., 2019). A multitaxon approach will be of great help for assessing with greater precision the patterns of biodiversity loss and the environmental factors that determine the response to these disturbances (Decaëns et al., 2018;Püttker et al., 2020). Such information is of great importance for understanding the drivers of the impacts of habitat loss (Carrié et al., 2017;Filgueiras et al., 2019a). ...
Article
Full-text available
Introduction: The effects of habitat transformation have been widely studied and the effects are well-known at different levels of biological organization. However, few studies have focused on responses to this process at the level of multiple taxa in diverse taxonomic and functional groups. Objective: Determine the variations in taxonomic and functional diversity of ants, butterflies, and dung beetles, at a spatial and temporal level in a landscape mosaic of the ecoregion of the Colombian foothills. Methods: We assessed amount of natural habitat and landscape composition in four types of vegetation, during the highest and lowest rain periods. We collected butterflies with hand nets and used baited pitfall traps for dung beetles and ants. Results: Habitat loss positively affected ant and butterfly species richness, and negatively affected dung beetles. The abundance of ants and butterflies had a positive effect on the dominance of species in the transformed vegetation , for dung beetles the abundance was negatively affected by the absence of canopy cover. Habitat loss had no negative effect on functional diversity as there is no difference between natural and transformed vegetation. Conclusions: The amount of habitat, habitat connectivity and different types of vegetation cover were important factors in the maintenance of insect diversity in the modified ecosystems of foothills of the Colombian Orinoquia. The lack of a common spatial and temporal pattern shows that studies of multiple insect taxa should be carried out for biodiversity monitoring and conservation processes.
... It should be noted that along with the loss of cultivated land area, the degree of fragmentation of China's cultivated land is increasing. This can not only cause a decrease in the value of ecosystem service functions and biodiversity, but can also exacerbate the inefficient use of cultivated land and have a negative impact on national food production [28][29][30][31]. ...
Article
Full-text available
Cultivated land area and fragmentation are two crucial elements that influence food production in China. However, few studies have analyzed the fine-grained changes in both the area and the fragmentation of cultivated land from a national perspective. Using a 30 m annual land-cover dataset with the help of cultivated land area (CA) and a newly constructed coupling landscape fragmentation index (CLFI), this research identified the spatial distribution characteristics of both the area and the landscape fragmentation of cultivated land in Chinese counties and simultaneously analyzed the changes in the above two elements since the beginning of the 21st century. The results showed that the distribution pattern of CA in China’s counties was large in the north and small in the south. During the first two decades of the 21st century, the total CA of China decreased, but the decrease rate decreased from 2.91% in 2000–2010 to 0.41% in 2010–2020; in contrast, the CLFI of China showed an increasing trend, but the increase rate also decreased, from 4.06% in 2000–2010 to 0.51% in 2010–2020. Spatially, from 2000 to 2010, there was an obvious reverse distribution pattern between counties with large CA changes and those with large CLFI changes; however, this pattern became less obvious from 2010 to 2020. Urban expansion, sloping land conversion programs and other ecological protection programs, and land development and consolidation programs were major causes of changes in the CA and CLFI across China. The findings of this research provide a data basis and scientific support for the protection and large-scale utilization of cultivated land, where this information is necessary to ensure food security.
... Peregrine species with world-wide distributions replace the many local species that are unable to compete successfully in such altered environments. Termite communities also suffer from simplification associated with deforestation, whereas ant communities seem better able to adapt to disturbances globally (Decaëns et al., 2018;Lawton et al., 1998). ...
Article
Full-text available
Macroinvertebrates comprise a highly diverse set of taxa with great potential as indicators of soil quality. Communities were sampled at 3,694 sites distributed world‐wide. We aimed to analyse the patterns of abundance, composition and network characteristics and their relationships to latitude, mean annual temperature and rainfall, land cover, soil texture and agricultural practices. Sites are distributed in 41 countries, ranging from 55° S to 57° N latitude, from 0 to 4,000 m in elevation, with annual rainfall ranging from 500 to >3,000 mm and mean temperatures of 5–32°C. 1980–2018. All soil macroinvertebrates: Haplotaxida; Coleoptera; Formicidae; Arachnida; Chilopoda; Diplopoda; Diptera; Isoptera; Isopoda; Homoptera; Hemiptera; Gastropoda; Blattaria; Orthoptera; Lepidoptera; Dermaptera; and “others”. Standard ISO 23611‐5 sampling protocol was applied at all sites. Data treatment used a set of multivariate analyses, principal components analysis (PCA) on macrofauna data transformed by Hellinger’s method, multiple correspondence analysis for environmental data (latitude, elevation, temperature and average annual rainfall, type of vegetation cover) transformed into discrete classes, coinertia analysis to compare these two data sets, and bias‐corrected and accelerated bootstrap tests to evaluate the part of the variance of the macrofauna data attributable to each of the environmental factors. Network analysis was performed. Each pairwise association of taxonomic units was tested against a null model considering local and regional scales, in order to avoid spurious correlations. Communities were separated into five clusters reflecting their densities and taxonomic richness. They were significantly influenced by climatic conditions, soil texture and vegetation cover. Abundance and diversity, highest in tropical forests (1,895 ± 234 individuals/m2) and savannahs (1,796 ± 72 individuals/m2), progressively decreased in tropical cropping systems (tree‐associated crops, 1,358 ± 120 individuals/m2; pastures, 1,178 ± 154 individuals/m2; and annual crops, 867 ± 62 individuals/m2), temperate grasslands (529 ± 60 individuals/m2), forests (232 ± 20 individuals/m2) and annual crops (231 ± 24 individuals/m2) and temperate dry forests and shrubs (195 ± 11 individuals/m2). Agricultural management decreased overall abundance by ≤54% in tropical areas and 64% in temperate areas. Connectivity varied with taxa, with dominant positive connections in litter transformers and negative connections with ecosystem engineers and Arachnida. Connectivity and modularity were higher in communities with low abundance and taxonomic richness. Soil macroinvertebrate communities respond to climatic, soil and land‐cover conditions. All taxa, except termites, are found everywhere, and communities from the five clusters cover a wide range of geographical and environmental conditions. Agricultural practices significantly decrease abundance, although the presence of tree components alleviates this effect.
... Agriculture often involves large-scale shifts in nutrient flux and plant productivity and is a major driver of biodiversity change and habitat (Marta et al., 2021). The growing need for food production and energy resources continues to increase the pressure to expand agricultural lands, such that "agricultural frontiers" are now reaching the last unprotected natural areas in many regions of the world (Hubert et al., 2010;Decaëns et al., 2018). In the Eastern Plains (Llanos) of Colombia, diverse agricultural activities have exerted growing pressure on natural ecosystems during the last 50 years (Romero-Ruiz et al., 2012). ...
Article
Analyzing the impacts of agricultural activities on biodiversity requires a full knowledge of the ecology of the studied species. Using a trait-based approach may improve our ability to understand land use impacts on less well-studied species in order to establish general trends that will facilitate the prediction of these impacts. In this study, we applied a trait-based approach to understand the impact of land use change on ant communities in the Colombian Llanos region. Five common land uses were considered (annual crops, rubber plantations, oil palm plantations, improved pastures and semi-natural savannas) to test whether some morphological traits respond to soil properties and land uses. An RLQ analysis shows a significant common structure between species distribution, environmental factors, and morphological ant traits. This indicates that morphological traits could be used to predict the response of ant communities in different land uses since they respond to environmental characteristics as vegetation complexity, composition and management. Based on the selected morphological ant traits, three groups of land uses were differentiated: grazing-based systems, agroforestry plantations and annual crops. Agroforestry plantations, especially rubber plantations tend to host larger and specialized ant species, while grazing-based systems (i.e. improved pastures) mainly host small and generalist ants, and annual crops host more pigmented ants. These findings suggest that certain morphological traits can reflect the ability of ant species to settle down and survive in a given land use system. Our study shows that improving knowledge about trait-environment associations could be a useful way to better understand how ecological filtering shape neotropical ant communities and how they respond to landscape transformation and land use changes.
... No meado do século 20, o reconhecimento e a divulgação da crise ambiental revelaram as contradições da sociedade (PORTO; PORTO, 2015;SANTOS, 2015;SOUTO, 2016), formuladas em desigualdade social (BARROS et al., 2000;GENNADIEVNA et al., 2017), analfabetismo (BRAGA;MAZZEU, 2017;MORAES;SCHWENGBER, 2017), na poluição e na contaminação de rios (MEKONNEN; HOEKSTRA, 2017;WEN et al., 2017;ZHOU et al., 2020), no desmatamento e na perda da biodiversidade (CARDIL et al., 2020;DECAËNS, et al., 2018), na Oliveira Junior, Pereira, Lobão | 3 ...
Article
Full-text available
Por meio deste estudo, objetivou analisar o uso e cobertura da terra para fundamentar a discussão sobre a sustentabilidade e a gestão ambiental no município de Muritiba-BA. A área de estudo está inserida no bioma Mata Atlântica e no território de identidade Recôncavo, onde são desenvolvidas atividades agropastoris e urbanas. Mediante a interpretação da imagem do sensor Pleiades-1A, com resolução espacial de 50cm, realizou a vetorização em tela das classes e produziu um mapa de uso e cobertura da terra (1/50.000). Para tanto, seguiram as indicações do sistema de classificação do IBGE para vegetação e o uso e cobertura da terra, acordado com os dados verificados em campo, resolução da imagem e escalas geográficas e cartográficas. Os resultados demonstraram a preponderância da agropecuária, que recobrem o município com lavoura-pastagem alternadas. Essa atividade, associada às urbanas, suprimiram as formações vegetais primárias. Hoje, ocorrem pequenas e isoladas manchas de vegetação secundária, caracterizadas pela predominância da Mimosa tenuiflora (Mart.) Benth, da família Fabaceae-Mimosoideae, conhecida como Jurema. Constatou-se a dificuldade da gestão ambiental, devido às evidências da insustentabilidade e da materialização da crise ambiental, observada na supressão da vegetação da área de preservação permanente (APP), o que demonstra a incompatibilidade entre uso e legislação vigente.
... La deforestación derivada de ello generó impactos dramáticos en la riqueza de especies debido a que esta región alberga el 10% de los animales que habitan la tierra (National Geographic, 2019). Estudios anteriores determinaron que la deforestación produce una pérdida de entre 30 y 40% de la riqueza de las especies (Decaëns et al., 2018). Además, los impactos para la fauna y ora se presentan tanto inmediatamente como en el largo plazo, teniendo en cuenta que los bosques tropicales son necesarios para mantener el equilibrio ecosistémico (UNEP -UN Environment Programme, 2019). ...
Article
Full-text available
Este estudio muestra la responsabilidad que la multinacional JBS S.A., un gran productor de proteína animal en el mundo, ha tenido sobre la pérdida de biodiversidad por deforestación en Brasil, por más de diez años. Para ello se construye una contracuenta basada en información de diversas ONG y de organizaciones sociales que evidencian cómo los reportes de sostenibilidad de las grandes compañías maquillan comportamientos que llevan a la insostenibilidad teniendo como punto de referencia una perspectiva teórica crítica de la contabilidad y tomando como base puntual la literatura que aborda las posibilidades de la contabilidad para la emancipación, en diálogo con literatura latinoamericana sobre temas ambientales y del desarrollo. El trabajo se basa en teorías emergentes bajo el concepto de contracontabilidad, que plantean la débil efectividad del desarrollo sostenible. Los hallazgos de esta investigación evidencian las posibilidades de nuevos discursos contables para la transformación social y ambiental en los países latinoamericanos a través de maneras alternativas de rendición de cuentas, ampliando la literatura contable sobre el rol de la contabilidad en la problemática ambiental, en tanto los informes tradicionales y oficiales sólo muestran la cara conveniente a los intereses de la empresa que reporta.
... A floresta tropical brasileira da Amazónia está sob pressão desde a década de 1970, com a inauguração da autoestrada transamazónica a marcar o início da área de desflorestação «moderna» [2]. A perda de coberto da floresta amazónica nas últimas cinco décadas, tanto no Brasil como nos outros sete países, tem graves consequências para a região, nomeadamente a perda irreversível da biodiversidade [3], a degradação do solo [2], o menor volume de precipitação [4], o aumento das temperaturas atmosféricas [5] e uma maior suscetibilidade aos incêndios florestais [6], provocando ainda alterações climáticas significativas à escala regional, como a redução da precipitação na Amazónia meridional brasileira [7], a bacia do Prata [8], bem como à escala mundial [9]. As regiões do Sul, Sudeste e Centro-Oeste do Brasil, que produzem a maior parte das culturas do país, poderão sofrer uma perda de um bilião de dólares nos próximos 30 anos, devido à redução da precipitação resultante da degradação florestal contínua e da perda permanente de floresta na Amazónia [10]. ...
Technical Report
Full-text available
Este relatório comunica as estatísticas de desmatamento e degradação florestal 2002-2020 para a floresta tropical nos países sul-americanos da região amazônica, com base no novo conjunto de dados JRC Tropical Moist Forest (JRC-TMF). Além disso, o relatório descreve a dinâmica do desmatamento e degradação florestal na região, enfatizando vários tipos de degradação florestal e os efeitos da mudança da cobertura florestal relacionados à construção de estradas, áreas protegidas e disseminação de doenças zoonóticas.
Article
Full-text available
The Amazon basin is one of the most extensive, biodiverse, and dynamic tropical forest ecosystems on the earth. The Colombian Amazon basin occupies an area of approx. 34 million hectares, located in the country's southeast. The literature about the economic valuation of ecosystem services (ES) and the spatial information on remnant natural resources in the Colombian Amazon basin was revised through various information sources to document the earliest approximation to the state, spatial distribution, and economic value of the remnant natural capital at the scale of biomes, specific ecosystems, and political-administrative units. Our assessment estimated a natural capital loss of 18.1 billion $/year (equivalent to 6.7% of Colombian GDP in 2020) and a remnant natural capital worth 153.9 billion $/year (57 % of Colombia's GDP in 2020) for eight ecosystem services. This research founds that a potential expansion in extensive and intensive livestock production systems (in a ten-year projection) will generate an additional loss of remnant natural capital of approximately 9.1 billion $/year. Finally, considering that 63% of the remnant natural capital is represented in indigenous reservations and 28% in protected areas, it is essential that the political management of the Amazon Basin concentrate on strengthening these land management figures. Improving the governability of indigenous lands and incentive sustainable agricultural and cattle ranching production, are the principal political challenges.
Article
According to the habitat amount hypothesis the species richness in a sample site should increase with the habitat amount in the local landscape around the sample site. On the other hand, size and isolation of the patch containing the sample site should have no effect on species richness if the habitat amount in the local landscape remains constant. We tested these predictions using a small mammal dataset recorded in 100 transects in a savanna region of Brazil. We used generalized linear models considering the predictors - habitat amount, patch size, and patch isolation - together and singly, using subsets of data to control for correlations between them. We measured the habitat amount in 10 nested circles ranging in radius from 250 m to 6000 m around each transect. The scale of effect was the radius with the strongest estimated correlation between habitat amount and species richness. Our results indicate that habitat amount is the most important single predictor of species richness. The habitat amount effect was always positive and large. In general, patch size and patch isolation had no effect on species richness after controlling for the effect of habitat amount. The results are consistent with the habitat amount hypothesis, and suggest that habitat amount can be used in place of patch size and isolation in predicting small mammal species richness. This simplifies information needs by landscape managers as it replaces two predictors with one. Our results also support the use of estimated dispersal distance as a way of determining the appropriate landscape scale of effect.
Article
Despite a 79% reduction in deforestation rates since 2004 in the Brazilian Amazon, the current gross forest loss of 6207 km2 year−1 remains significant, promoting continuous habitat fragmentation. This study assessed the status of forest fragmentation in 2014 across the Brazilian Amazon applying a Morphological Spatial Pattern Analysis (MSPA). Our analysis provides the first 60-m-resolution fragmentation map for the entire Brazilian Amazon, including not only the commonly evaluated fragmentation classes such as core, edge and islet, but also four specific metrics measuring the effects of fragmentation on forest patterns (perforation, bridge, loop and forest branches). Areas of intense forest fragmentation were detected not only along the “arc of deforestation” of the eastern and southern Brazilian Amazon but in other regions as well. Considering a 1020-m edge width, we found that 28% (~891,000 km2) of the remnant forest was in non-core classes. Among these classes, bridge (i.e., forest corridors too narrow to contain core area) was the most representative (10.8%). The proportion of core area varied greatly among states, from 82.8% of the remaining forest in Amazonas to 26.1 and 13.5% in Maranhão and Tocantins, respectively. The detailed diagnosis of tropical forest fragmentation is a critical step for proposing solutions to maintain the stability of tropical forest services, including carbon storage, water and nutrient cycling, and biodiversity.