Biodiversity and Conservation

https://doi.org/10.1007/s10531-020-01999-3 (0123456789().,-volV)(0123456789().,-volV)

ORIGINAL PAPER

Discolouring the Amazon Rainforest: how deforestation is

affecting butterfly coloration

Ricardo Luı́s Spaniol, et al. [full author details at the end of the article]

Received: 10 October 2019 / Revised: 2 May 2020 / Accepted: 25 May 2020

Ó Springer Nature B.V. 2020

Abstract
Butterflies are among the most colourful organisms in the world and colour plays a central
role in many of their life-history strategies. However, the efficacy of coloration strategies
in these and other animals could be affected by sudden environmental changes, including
anthropogenic disturbances such as habitat loss and fragmentation. Here we investigate the
effect of forest disturbance gradients on the colours of fruit-feeding butterflies in the
Amazon Rainforest. The disturbance gradients tested represented habitat-size (continuous
primary forests versus forest fragments of 1, 10 and 100 ha) and succession gradients
(continuous primary forests, 30 year old secondary forests, and three year old early suc-
cessional forests). Using digital image analysis, we obtained intrinsic measures of butterfly
colour patches corresponding to hue, saturation, brightness, in addition to measures of the
contrast among patches and of the overall wing-colour diversity corresponding to 220
individuals, belonging to 60 species. Our results showed that butterflies in the secondary
forest and continuous primary forest are more colourful than those found in early suc-
cession and fragments of forests. Individuals occurring in forests of early succession
showed higher average values of hue and saturation, but lower brightness. Accompanying
changes in colour composition, wing-colour diversity among species was lower in human-
disturbed habitats, such as those of early forest succession and secondary forest. Forest
fragments have smaller effects on butterfly colour composition, indicating that well-
structured forest habitats can house butterfly assemblages with more diverse phenotypic
features and colours. We show how high deforestation rates in recent years is linked with
negative changes in functional coloration strategies (e.g. camouflage, warning colours),
something that has to date been poorly explored or demonstrated for butterflies. Specifi-
cally, human interference has apparently placed butterflies under strong selection for lower
diversity in their colours and range of defensive strategies. Those species that are most
colourful are the first to be locally extinguished, likely due to removal of native vegetation
and increased exposure to predators, and more broadly owing to

Communicated by Akihiro Nakamura.

This article belongs to the Topical Collection: Forest and plantation biodiversity.

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10531-020-

01999-3) contains supplementary material, which is available to authorized users.

123
 Biodiversity and Conservation

inhospitable environmental conditions. This illustrates an accelerated loss of local fauna

and a ‘‘discolouration’’ of the Amazonian butterflies due to anthropogenic impacts.

Keywords Anthropogenic gradients  Colour diversity  Conservation  Forest succession 

Fruit-feeding butterflies  Tropical forest

Introduction

Research on animal coloration is a classical issue that has long fascinated naturalists,
including back to the great expeditions of influential ones during the 1800s (Bates 1862).
This interest is partly due to the relevance of colours for understanding intra- and inter-
specific interactions, links to the abiotic environment, and the importance of such traits in
illuminating our understanding of ecological and evolutionary processes (Stevens 2016;
Endler and Mappes 2017). Through coloration, animals display broad repertoires in anti-
predator and mating strategies, aiming to maximize survival and reproductive success
(Cuthill et al. 2017).
Regardless of the specific strategies (e.g. anti-predatory or reproductive) employed by
an organism, environmental characteristics work as key aspects selecting phenotypic
patterns (Roslin et al. 2017). For example, recent research continues a long tradition in
testing how visual signal transmission is related to the habitat in which animals occur and
their contrast with the visual environment (e.g. Cheng et al. 2018; Willmott et al. 2017;
Walton and Stevens 2018). Therefore, environmental changes are expected to impact the
adaptive value of individual coloration (Delhey and Peters 2016), in the same way that
species assemblages from different habitats may house individuals with specific functional
features and adaptations (Spaniol et al. 2019).
In butterflies, their evolutionary history and current selection pressures acting on phe-
notypic variation are printed on their wings, ranging from conspicuous colours or cam-
ouflage, to wing-eyespots and mimetic complexes (rings), which together are responsible
for colour diversity in nature (Joshi et al. 2017). Environmental changes can affect
colour patterns on butterfly wings in different ways. One of these is through the supply of
host plants as food resources to immature individuals (Talloen et al. 2004). Caterpillar diet
is often essential for chemical compound synthesis (e.g. flavonoids), which leads to diverse
phenotypic expression, including of body and wing colours (Johnson et al. 2014). In
addition, abiotic conditions may change wing patterns along different microhabitat gra-
dients (Papageorgis 1975). Given that resource availability is linked to the abiotic con-
ditions of each habitat, once plant community species composition is modified, more
specialized butterflies (in terms of food resources) may disappear (Soga et al. 2015), or
differ in colour pattern due to the nature and amount of available resources. A second way
to understand the process of evolution and adaptation in butterfly colours consists of
knowing how visual signals are transmitted when natural habitats are modified. Different
communication strategies, including those used to mediate prey-predator and potential
mating interactions may be affected by structural changes in the environment (Briolat et al.
2018; Dalrymple et al. 2018). Abiotic components (e.g. temperature, humidity, brightness),
besides affecting animal thermoregulation, also have important roles determining energy
and resource inputs in ecosystems (Dalrymple et al. 2018), impacting factors such as
vegetation structure, visual backgrounds, and the light environment. In turn, in environ-
ments modified by anthropogenic impacts, the visual environment for signalling or against
which animals would normally be hidden may change. This will affect the efficacy of

123
Biodiversity and Conservation

colour strategies, either those relying on visual conspicuousness (e.g. mating and warning
signals), or on concealment, affecting fitness.
In tropical forests, high biological diversity can lead to the coevolution of several
adaptive strategies, including butterfly interactions with predators, competitors, mutualists,
or even potential mates. This makes the tropical region not only rich in species but also in
their phenotypic and behavioural diversity, including colour related strategies (Adams
et al. 2014). The Amazon Rainforest is an ideal region to test the above issues because it is
still largely preserved, yet is under great threat and suffering extensive deforestation,
affording the potential to study organismal responses to human-induced environmental
disturbance (Mesquita et al. 2015). In recent decades, this region has been widely trans-
formed with high deforestation rates driven by the advance of agricultural frontiers, with
concerning consequences for tropical biodiversity (Vieira et al. 2019). At present, public
policies and interventions adopted are sparse and have promoted insufficient results for
controlling the loss of huge areas of forest every year (Arima et al. 2014), which is
accompanied by high species extinction rates (Stork 2010; Barlow et al. 2016). The
immediate consequences of forest transformation are often unpredictable and require
careful monitoring of ecological and evolutionary responses from different species (Caro
et al. 2017).
Butterflies represent ideal study models for environmental assessments and allow rel-
atively easy monitoring in nature (Freitas et al. 2014). These insects manifest a large
repertoire of colours, which may be associated with the different habitats they occupy and
different strategies for distinct signalling functions (Endler 1993; Dalrymple et al. 2015).
Because butterflies play a diversity of ecological roles within complex networks, their
responses can be extended to evaluate the effects of environmental degradation in the
tropics (Spaniol et al. 2019). The fruit-feeding guild, in particular, comprises approxi-
mately 50–75% of all butterfly species belonging to the family Nymphalidae found in the
Neotropics. This group is represented by four subfamilies: Charaxinae, Biblidinae,
Satyrinae, and also a few tribes within Nymphalinae (Brown 2005). From the perspective
of different coloration properties (hue, saturation, brightness, wing-colour diversity), our
study aims to advance knowledge on how deforestation and habitat-size decrease may
influence the colour patterns observed in those species that remain, and the potential effects
on their anti-predator strategies. Environmental selection for specific colour combinations
in species assemblages would be less intense in the primary forest, given that a variety of
strategies and colour types should be able to thrive here with a greater range of visual
backgrounds and light conditions. This should be reflected in greater colour diversity in
more intact forest habitats, owing to a higher species richness and greater butterfly
assemblages. In comparison, in forests of greater levels of fragmentation and earlier
succession, fewer colour strategies may thrive, reducing overall species diversity, and this
may be especially true for phenotypes that rely on specific linkages between colour phe-
notype and visual background or light conditions.
We analysed the coloration of butterflies found in different forest patches of different
human impact, and classified butterflies based on putative defensive strategies (crypsis,
transparency, warning signals, and eyespots). We hypothesized that fruit-feeding butterflies
along anthropogenic gradients would be distinct with regards to their colour patterns, and
that prevalent phenotypic characteristics should change according to habitat size and
regeneration stage after a disturbance, including changes in the visual environment (light
conditions, background contrast) (Fig. 1). From this viewpoint, we expect that: (i) with a
decline of vegetation structure and changes in abiotic characteristics, as well as homog-
enization due to human activities will result in conspicuous butterflies being removed first,

123
 Biodiversity and Conservation

while broadly cryptic individuals (mainly brown winged butterflies) may persist by
avoiding detection against generally brown substrates along the gradient. (ii) Colourful
butterflies may be those most affected by forest disturbances, being the first to disappear
due to increased exposure in more open habitats. This way, colour diversity and richness
should be lower in disturbed habitats when compared to the preserved primary forest. As a
counter-prediction, some types of intrinsically conspicuous coloration may be effective to a
certain extent regardless of the visual environment (e.g. aposematism) and therefore persist
even in degraded habitats. In contrast, specialist species with camouflage that is

123
 Biodiversity and Conservation

b Fig. 1 Conceptual framework and predictions on colour-habitat patterns for fruit-feeding butterfly
communities in the Amazon Rainforest. From the regional species pool, we expect that butterfly
assemblages under similar habitats in the intact forest can have a higher colour composition compared to
degraded habitats, where selection pressures on phenotypes may change and select for fewer colour patterns.
Under human-induced changes, anti-predator strategies may display different performances, leading to
differences in some colour properties. Colour richness and intensity (saturation), differences between colour
patches, and wing colour diversity may decrease with a greater homogenization. In forests made more open
and brighter, conspicuous colours that are highly contrasting with altered backgrounds may not have the
same performance due to increased exposure of those butterflies to threats. Camouflage involving a
resemblance to the remaining substrate after a disturbance, combined with the presence of wing-eyespots,
may increase the survival chances of some butterflies, avoiding detection or directing predator attacks to less
vital parts of the body. Overall, we expect a shortage of colourful butterflies associated with deforestation in
the Amazon Rainforest, so that the Amazonian fauna is currently undergoing a process of discolouration

specifically effective on a limited range of backgrounds may decline if those backgrounds

are removed.

Materials and methods

Study area

Butterflies were sampled in areas of the Biological Dynamics of Forest Fragments Project
(PDBFF) (2° 210 36.1400 S, 59° 570 45.6000 W), belonging to the National Amazon Research
Institute (INPA). These are spread over 1000 km2, 90 km north of Manaus, Amazonas
State (AM), in a Brazilian Federal Protected Area created in the late 1970s to investigate
the consequences of deforestation and forest fragmentation in the Amazon (Biereegaard
et al. 1992). Three farms (Dimona, Porto Alegre, and Esteio) were defined as study areas.
Each study area includes the following habitats: forest fragments of 1, 10, and 100 hec-
tares, and habitats varying in succession, from an initial stage of succession around the
previous fragments, to secondary forests, and large extensions of primary continuous forest
(Fig. 2). The initial succession comprised vegetation with approximately three years of
regeneration, and isolated fragments from the other habitats. The secondary forest covers
areas formerly used as cattle pasture, and which now have forests with 30 years of
regeneration. Forest fragments are isolated patches of primary forest, structurally similar to
the large areas of continuous forest (Laurance and Vasconcelos 2009). The primary con-
tinuous forest occupies extensive adjacent regions, and is used in this study as a control. In
these landscapes, two distinct environmental gradients can be identified and monitored to
understand the response of organisms to habitat variations: habitat-size and forest suc-
cession. We refer to the forest succession gradient as including the areas in initial suc-
cession, secondary forest and continuous forest; the habitat-size gradient, in turn, includes
fragments of 1, 10 and 100 hectares, and the continuous forest as a control.

Data sampling

Two field expeditions were performed between August and September 2015 and 2016, at
the beginning of the dry season in the Amazon region. Each farm received a total of 12
sample units (SUs), each of them with five portable traps, containing attractive bait made
with bananas fermented in sugar cane juice for 48 h (Freitas et al. 2014). Sample units
were installed on each farm, with three SUs in early successional sites, three in secondary
forest sites, one for each forest fragment size (1, 10, and 100 hectares) and three points in

123
 Biodiversity and Conservation

Fig. 2 Map with the three study areas at the Biological Dynamics of Forest Fragments Project (PDBFF),
indicating the location of Dimona, Porto Alegre and Colosso farms. White patches represent deforested
areas initially used for grazing purposes. The green portion corresponds to primary intact forest areas,
including all the forest fragments used for the experiments in each study area (1, 10, and 100 hectares). The
inset reveals the habitat disposition on each farm: green is for primary forest (either continuous or in
fragments), white for secondary forest (30 years of regeneration), and black for early succession patches
(three years after a disturbance—around fragments)

the adjacent continuous forest, totalling 60 traps per farm. A minimum distance of 20 m
between neighbouring traps was observed, placed between 100 and 130 cm above the
ground (Freitas et al. 2014). SUs were kept at least 500 m from each other within each
farm.
At each sampling occasion, traps remained exposed for eight consecutive days in each
farm, being reviewed at 48 h intervals to renew the bait, collect, identify, mark and release
the captured butterflies. During the review process, two samplers equipped with entomo-
logical nets performed active samples of fruit-feeding butterflies around the traps, as long
as they were visibly attracted by the bait. These combined methods allow us to comple-
ment the representation of butterfly species, including especially those species attracted but
not caught by the traps (Checa et al. 2018). At least 10 individuals of each species were
collected for subsequent lab measurement, and about three to four individuals per species
were included, on average, for colour analysis (Fig. 3). Whenever possible, we selected the
same number of males and females per species to account for sexual dimorphism.

Colour measurements

The collected butterflies were fixed and deposited in an entomological collection (Labo-
ratório de Ecologia de Interações—UFRGS) for gathering individual information. From 60
fruit-feeding butterfly species, we selected 220 individuals with well-preserved colour
characteristics. Under natural light conditions, each individual was photographed in a
ventral position. We took all pictures using a Nikon D5300 camera with 18–55 mm lens,
accompanied with a scale bar and a grey card (18%) to correct for variation in light
conditions among photographs (Stevens et al. 2007). The photographs were taken in RAW
format and selected for appropriate exposure in RawTherapee software (version 5.3). All

123
Biodiversity and Conservation

Fig. 3 We measured the colours of 220 individuals belonging to 60 species of fruit-feeding butterflies.
a Transparent butterfly (Cithaeria andromeda) with camouflaged wings. Photograph by Iserhard CA. b Owl
butterfly (Caligo teucer) with typical wing eyespot. Photograph by Spaniol RL. c The colourful Prepona
narcisus with highly contrasting wings. Photograph by Spaniol RL

images were imported into the ImageJ program and through the MICA Toolbox add-on
(version 1.22), and we generated multispectral images ‘‘.mspec’’ calibrated from the grey
card (Troscianko and Stevens 2015). For every image, we drew ‘‘regions of interest’’
(ROI’s) around the wing areas that we wished to measure, based on the main colours found
on the wings. ROI’s were chosen based on identifying colour spots that could transmit
visual information, such as wing areas with strong colour contrast with the wings as a
whole. All photographed butterflies had their colour patterns quantified following a pre-
vious approach that made use of calibrated digital images (Stevens et al. 2007).
From the standardized images, we obtained colour data corresponding to reflectance in
the three camera colour channels: red, green, and blue (RGB) on a scale of 0–255 for each
colour channel. From these values, we were able to calculate several colour variables for
each butterfly colour patch: hue (the colour type, e.g. orange or blue), saturation (colour
‘richness’ or intensity when compared to white light), brightness (the sum of the reflec-
tance values along the spectral range 300–700 nm), maximum contrast between patches
(the greatest Euclidean distance measured between two patch colours on an individual in a
colour space; see Data Analysis below), and wing-colour diversity (diversity of colours
achieved on a single individual) (see Stevens et al. 2007). Finally, we present the frequency
of four main putative anti-predator strategies: camouflage, transparency, warning colours,
and wing eyespots. Each species and its type of defence was categorically classified by
their general composition and likely interaction with the visual environment, observing the
predominance of brown/cryptic colours, translucent wings, highly contrasting colours (for
example, yellow, red, orange combined with black or white), and circular features on the
wing resembling an eye. Brown/cryptic colours are categorized based on ecophysiological
(thermoregulation) and defence (predation and detection avoidance) features. These colour
patterns provide evidence on the habitat used by the species characterized by dark or light

123
 Biodiversity and Conservation

coloration, in which dark colours are essential for both camouflage and thermoregulation
(Kingsolver 1985) in forested habitats, since dark colours increase radiant heat absorption.
On the other hand, light colours are expected to be found in sites in which overheating is
more likely (Ellers and Boggs 2004), leading to higher chances of desiccation. Despite
contrasting information about the functional role of eyespots, they can be characterized by
the combination of ocellus on the ventral surface of wings, and have one of two anti-
predator functions: (i) deflection, whereby a predator attack is redirected to non-vital body
parts of the insect (many small ocellus occur on the border of the hind wing on the ventral
surface) (Stevens 2005); (ii) intimidation, in which the pattern of the wings may imitate a
dangerous organism that is a threat to the predators themselves, such as the mimicry of owl
eyes, or simply by generating a conspicuous display (one big spot on the hind wing in the
ventral surface) (Stevens and Ruxton 2014). We appreciate that this is to a certain extent
subjective but full categorisation of the type of defence each species primarily relies on
methods beyond the scope of this study (requiring, for example, toxicity analysis and
behavioural experiments).

Data analysis

We converted standardized RGB reflectance values to XY coordinates in a triangular

colour space, whereby the centre of the space represents the achromatic point and the
location around this corresponds to a given colour type and saturation. This approach is a
common method used to characterise the nature of colour variation that exists, and quantify
aspects of colour such as saturation and contrast, given a certain dimension of colour space,
here based on three colour channels (see Stevens et al. 2009). We calculated saturation as
the Euclidian distance from the centre of the colour space, whereby greater distances
equate to higher levels of colour richness (Endler and Mielke 2005; Stevens et al. 2014).
To calculate hue, we followed a range of past approaches in describing hue in the form of
colour channels that describe the variation in colours present in the dataset (Komdeur et al.
2005; Spottiswoode and Stevens 2012; Stevens et al. 2014). This is not intended to directly
mimic how visual systems process colour types with specific opponent colour channels, but
is broadly similar in nature. To determine the most appropriate channels, a principal
component analysis (PCA) was used to summarize the importance of each colour channel
in butterfly phenotypic variation and to define the colour types that exist along relevant
axes of variation in colour space. To eliminate achromatic variation and to analyse only
colour differences, we removed brightness from the dataset by standardizing the colour
values to proportions (Cuthill et al. 1999; Spottiswoode and Stevens 2012), and then from
the resulting PCA derived colour channels that best described the variation in colour that
existed among species and colour patches (Spottiswoode and Stevens 2012; Stevens et al.
2014). These colour channels define types of colour in the form of a ratio, which corre-
sponds to an axis of variation running through the colour space. To quantify wing-colour
diversity, we use the mean Euclidean distance between x, y colour space coordinates for all
individuals within each treatment. Through a resampling analysis (random sampling with
replacement) we corrected the differences in sample size among habitats in each gradient.
The number of resampled butterflies was equal to the smallest group in the analysis (see
Spottiswoode and Stevens 2012). For brightness, we calculated the average brightness in
reflectance across the RGB channels.
All colour variables (hue, brightness, saturation, differences between patches, and
colour diversity) were then individually analysed along the habitat-size and forest

123
Biodiversity and Conservation

succession gradients, using generalized linear mixed models. For these analyses, we input a
colour variable (the response) and environmental data (the explanatory variables) as fixed
factors, and we also controlled for differences in abundance among species as a random
factor. In order to make the coloration results independent of the butterfly assemblages in
each habitat, we also controlled for the effects of species composition incorporating species
identification in the generalized linear mixed models, and running it as random factor. We
also present butterfly richness data for each habitat in a Venn diagram, and composition
changes across habitats using a non-metric multidimensional scaling (NMDS) ordination,
complemented with statistical testing by permutation via Adonis, using 999 permutations
and Bray–Curtis index. All computations were performed with R, using PAVO and lmer4
statistical packages (Maia et al. 2013; Bates et al. 2015; R Core Team 2018). Sample
coverage for each habitat was calculated to estimate the representativeness of fruit-feeding
butterflies through the q statistics as proposed by Chao et al. (2013). This analysis was
performer using the software iNEXT online (Hsieh et al. 2016).

Results

From the 60 species sampled in field, 30 species (95 individuals) were found in primary
forest, 33 (85) in secondary forest and 29 (229) in early successional habitats. Primary
forest and early successional assemblages harboured most unique species, but just one
species was common for both habitats only. Eight species were common to all habitats
across this gradient (Fig. 4), and several were found in at least two kind of habitats. In
general, there were important changes in butterfly composition among the successional
habitats (Fig. 5). Forest fragments (1, 10, and 100 hectares) housed 10 species (18 indi-
viduals), 11 (17) and 13 (27), respectively (Fig. 4). No significant changes were found in
assemblage composition among the forest fragments (S1). The representativeness of fruit-
feeding butterflies for each habitat as measured by sample coverage was respectively:
3 years = 0.97; 30 years = 0.87; Primary/Continuous Forest = 0.84, 01 hectare = 0.73, 10
hectares = 0.61, 100 hectares = 0.78.
We checked if there was structure in colour composition of the fruit-feeding butterfly
assemblages along both habitat-size and forest succession gradients. From the achromatic

Fig. 4 Venn diagram comparing species richness and shared species among habitats making up:
a the Sucessional gradient and b the Habitat-size gradient

123
 Biodiversity and Conservation

Early succesional
Secondary forest
0.2

Primary forest
0.1

Primary_forest
axis 2
0.0

Secondary_forest

Early_succesional
-0.1
-0.2

-0.4 -0.2 0.0 0.2 0.4

axis 1

Fig. 5 Non-metric multidimensional scaling (NMDS) ordination of Bray–Curtis distance matrices for
butterfly assemblage composition in samples from early successional habitats (3 years after a disturbance),
secondary forests (30 years of regeneration), and primary intact forest in the Amazon Rainforest. The
diagram was constructed by using the ‘ordispider’ function of vegan package to illustrate the habitat and
samples. Each dot represents the sample unit ensemble from one site. Statistical testing by permutation (999
times) via Adonis indicates differences in composition among the habitats: R2 = 0.14, P = 0.001

central point (grey) in the colour space triangles, we observed a trend for longwave (LW)
colours in all the habitats, and the colour values cross through the centre of the colour
space from long to shortwave (SW) and mediumwave (MW) parts (Fig. 6). This was
confirmed by Principal Component Analysis (PCA) using a covariance matrix with stan-
dardized values for each colour channel.
For the forest succession gradient, the first two axes of the PCA expressed 99% of all
variation in butterfly colour. The first axis separated LW from the other colour channels,

Fig. 6 Butterfly reflectance in a colour space triangle. Each dot represents the reflectance values for each
individual observed along the forest gradients in the Biological Dynamics of Forest Fragments Project
(BDFFP) areas: a Forest Succession; b Habitat-size. All dots were coloured according to the treatments
identified through the legend. S shortwave, M mediumwave, and L longwave refers to the wavelength
according to the RGB reflectance values

123
Biodiversity and Conservation

explaining 79% of the phenotypic patterns (i.e. LW/[SW ? MW]). Larger values corre-
spond to patches that are more red in colour, whereas lower values to patches that are more
blue or green. The second axis contained 20% of the variation, separating SW from LW
and MW (SW/[LW ? MW]: PC1 coefficients: SW: 0.554; MW: 0.525; LW: - 0.645.
PC2: SW: 0.660; MW: - 0.749; LW: - 0.042). Larger values correspond to blue colours,
and smaller values to those that are more yellow/brown. The first two axes of the PCA also
expressed 99% of the phenotypic variation of the butterflies to the habitat-size gradient. In
a similar pattern, the first axis separated LW from MW and SW, explaining 72% of the
phenotypic patterns observed. The second axis contained 27% of the variation, separating
SW from LW and MW (PC1 coefficients: SW: 0.553; MW: 0.483; LW: - 0.677. PC2:
SW: 0.634; MW: - 0.771; LW: - 0.031).
Significant changes were found along the forest succession gradient for the following
colour variables: hue, saturation, brightness, and wing-colour diversity. Higher hue values
were observed in primary forest regions (F value = 6.60, P = 0.01). In other words, longer
wavelength rich colours (LW) such as brown/red prevail in older forested areas regarding
succession. For saturation, we detected lower values in areas of early succession (F
value = 4.48, P = 0.03). This means that butterflies are less colourful in recently disturbed
habitats when compared with continuous forest. Brightness values were higher in the early
succession and secondary forest habitats (F value = 3.93, P = 0.05) (Fig. 5), in which
butterflies with lighter colours are more common. The forest interior may be housing
melanic patterns, phenotypically darker individuals. We found a larger wing-colour
diversity for the continuous forest (P \ 0.05 for 75.6% of resamples with N = 40) when
compared to the treatments under succession. A similar pattern was also observed for forest
patches (P \ 0.05 for 67.1% of resamples with N = 7), which showed a lower colour
diversity than continuous forest (Fig. 7). Larger forest habitats can house butterfly
assemblages with more diverse phenotypic features and colours. Boxplots of the non-
significant variables of colour for fruit-feeding butterflies are attached in the supplements
(S2).
Among putative anti-predatory strategies using colour patterns (Fig. 8), camouflage
appears as a predominant feature in all studied habitats. As the forest succession pro-
gresses, wing-eyespots become more important and appear frequently. Transparency seems
to be an exclusive strategy for the well-preserved continuous forest. Thus, the co-occur-
rence of all different phenotypic manifestations is observed exclusively under areas of
intact vegetation, and reinforces our results on colour diversity.

Discussion

In agreement with our predictions, the colour composition of butterfly assemblages is in

close association with habitat features. Several butterfly colour variables changed among
habitats, especially along the forest succession gradient. The butterfly species composition
followed the same patterns for this gradient, indicating segregation among butterfly
communities in initial stages of succession, secondary forests, and primary continuous
forest. Butterflies using conspicuous colour patterns seem to be the most affected by
human-activities, appearing among the first to disappear from recently disturbed forest
regions in the Amazon. On the other hand, this study also demonstrates the positive effect
of natural regeneration in recovering colour composition and ecological structure in
modified tropical forests (Crouzeilles et al. 2017). The maintenance of both primary and
secondary forests in human-modified landscapes might be helpful for re-colonization of

123
 Biodiversity and Conservation

Fig. 7 Boxplots of colour variables for fruit-feeding butterflies. Wing-colour diversity in the different
habitats in areas in the Biological Dynamic of Forest Fragmentation Project (BDFFP) of the a habitat-size
and b forest succession gradients. c Hue; d Saturation and e Brightness in habitats of the forest succession
gradient (Early Succession—3 years after a disturbance, Secondary Forest—30 years of regeneration and
Primary Continuous Forest)

plant species and interaction networks (Pellissier et al. 2018; Rozendaal et al. 2019). In the
same way, the regeneration of perturbed habitats over decades of succession may also
assist in the maintenance of species composition, and of the diversity of protective col-
oration and signalling in butterflies associated with forest interiors. Our results are

123
Biodiversity and Conservation

Fig. 8 Main defence strategies manifested by butterflies (camouflage, transparency, warning-colour, and
wing eyespots) for different habitats that make up the succession (a) and habitat-size (b) gradients in the
Amazon Rainforest (for details see text)

supported by the large sampling coverage calculated for all types of studied habitats in the
Amazon.
Whatever the kind of interaction, environmental features can produce effects on species
assemblages through individual colour composition, and this was observed for butterfly
response to forest degradation and habitats loss in the Amazon (Spaniol et al. 2019).
Firstly, camouflage appears as a dominant feature against predation in many animal groups
across the tropics, including birds and butterflies (Dalrymple et al. 2015). With a pre-
dominance of individuals with cryptic colours (especially browns) throughout the forest
succession gradient, including in continuous forest, this study supports that idea, sug-
gesting that camouflage is one of the main defence strategies for Lepidoptera and is a
valuable defence even as the habitat composition is altered. This also suggests that gen-
erally dull cryptic colour patterns may enable concealment in a range of habitats. More-
over, a lower colour saturation of butterflies occurring in recently disturbed sites reveals
that conspicuous individuals may be receiving greater predatory pressures, especially in
open habitats. While one may expect that conspicuous markings may operate across a
range of visual backgrounds, it is possible that these conspicuous species may become too
exposed when contrasting with new backgrounds created by disturbances. Consequently,
more colourful butterflies, with the exception of aposematic patterns, may be among the
first to disappear locally soon after the deforestation process. The above mentioned results
are corroborated with changes in beta diversity of fruit-feeding butterflies in the different
habitats of the successional gradient. In this case, species composition is directly linked to
changes in colour patterns in the Amazon region. In addition, habitat degradation may alter
the composition of predators and allow a greater number of naı̈ve predators into the forest
area (Bruno and Cardinale 2008; Ciuti et al. 2012). This may result in greater attack rates
on butterflies that normally can rely to a great extent on the learnt avoidance by predators
of their conspicuous warning, startle, or mimicry signals. From a high diversity regional
pool of species, selection may therefore lead to lower diversity in butterfly colours, as well
as in their protective and signalling strategies, in degraded habitats.
Furthermore, we note that conspicuousness is not necessarily an exclusive feature of
butterflies living in primary continuous forests. There is a remarkable variance in the
colour properties we analysed (hue, brightness, saturation and colour diversity), showing
that conspicuous and cryptic individuals even with patterns conferring different

123
 Biodiversity and Conservation

communication roles can make use of the same habitats. In addition, habitats in a preserved
forest are capable of harbouring greater colour diversity and subsequently more anti-
predator strategies. This may be possible because of the large supply of substrates such as
leaves, trunks, branches, stones, with which the individuals can ‘‘interact’’ in terms of their
phenotypic appearance (Pinheiro et al. 2016). Such environmental resources are not always
readily available after a severe disturbance, where abrupt and intense changes can lead to
modifications in the habitat structure and vegetation resilience trajectories (Jakovac et al.
2015). Differences may also be strongly controlled by changes in predator communities
(Mappes et al. 2014).
Analysing the habitat-size gradient, we discovered that forest patches with different
sizes might have smaller effects on butterfly assemblage colour composition. Our study
areas comprise an already well-developed forest matrix, which may be permeable to many
species, including forest specialists (Schtickzelle et al. 2007). Reduced forest patches do
not necessarily lead to strong changes in vegetation composition or structural features
(Brown and Hutchings 1997), which however are easily identified along the forest suc-
cession (Mesquita et al. 2015). Therefore, the supply of immature host plants, adult food
and substrates may remain. Complementarily, the way visual signals are propagated may
remain mostly unchanged since forest characteristics are substantially maintained. This
means butterflies with distinct phenotypes may be able to move among forest patches with
a relatively reduced risk of predation. Environmental conditions and resources such as
leaves, twigs, trunks and light entry through the forest canopy may also remain available
and relatively constant, helping colour patterns remain the same within forest habitats
patches.
Our findings suggest that butterflies from a regional species pool are under selection for
lower diversity in their colours as well as in their defence strategies due to human inter-
ference. The same has been observed for species richness and composition (Spaniol et al.
2019). Coloration is only one of several organismal traits offering quick responses to
environmental changes, but this has been seldom applied in our quest to understand threats
to biodiversity and processes that lead to species extinctions (Hook 1997; Caro et al. 2017).
We highlight the study of animal coloration as an excellent basis for the evaluation of
environmental health and in the planning for management and conservation of forest
ecosystems. Colour shifts are important markers of several anthropogenic changes (Caro
et al. 2017), and by observing phenotypic aspects in animals, we can systematically
monitor the performance of protective coloration and signalling under different environ-
mental conditions. Certainly, the colour patterns observed here are not only due to
increased predation but might have alternative explanations. Species composition changes
could be caused by many drivers aside from predation rates on adults mentioned above.
Changes could also be caused by decreased host plants, increased larval predation rates in
more opened forest structures, increased parasitoid abundances, or increased predator
populations (for both butterfly and caterpillar). Therefore, a next important step is to
investigate whether the reported pattern reflects ecological filtering or adaptation
processes.
At a time when deforestation rates are increasing again in the largest tropical forest in
the world (INPE 2019), this study draws attention to the importance of maintaining well-
preserved forest remains in the Amazon. In addition, forested habitats that have been
recovering for 30 years (secondary forest) show increasing in colour diversity, allowing the
movement and permanence of organisms near the primary forests, maintaining ecosystem
services. Thus, the maintenance of legal reserves inside private properties and restoration
programs in degraded areas with higher biological importance (Metzger et al. 2019; Vieira

123
Biodiversity and Conservation

et al. 2019) should be encouraged, being key factors aiming conservation of biodiversity in
tropical forests. We also suggest that the presented results and actions should be incor-
porated into public policy, reinforcing that information based on scientific research is
crucial to the decision-making process. Once unplanned, forest conversion leads to con-
stant uncertainties about the compliance with Brazilian environmental legislation, we
provide strong evidence that a significant portion of the butterfly fauna may be facing
disappearance from tropical forests in the near future. This is just the visible part of the
Amazonian fauna—in the literal sense of the word, where the most colourful butterflies are
the first to be locally extinguished by removing the native vegetation. It is necessary to use
pragmatism, and this insect group holds the public eye enabling effective warnings about
the limits of forest exploitation. Otherwise, the largest rainforest in the world will be less
colourful without some of its main inhabitants.

Acknowledgements We thank the Biological Dynamics of Forest Fragments Project (BDFFP) through the
Thomas Lovejoy research fellowship program for providing financial and logistical support for fieldwork.
This is the study 789 of the Biological Dynamics of Forest Fragments Project (BDFFP-INPA/STRI). We
also thank Yang Niu, Anna Hughes, Gabriel Nakamura and two anonymous referees for providing useful
comments that improved on the manuscript. Ricardo Spaniol thanks Capes for his PhD fellowship. Sandra
Hartz (Process 304820/2014-8) and Milton Mendonça (Process 309616/2015-8) are CNPq researchers.
Chico Mendes Institute for Biodiversity Conservation—ICMBio conceded licenses and permissions for the
research activities (License numbers 48786-1 and 48786-2).

Author contributions SRL, MMS and SM conceived the ideas and designed methodology. Fieldwork was
conducted by SRL and ICA. Image analysis and statistical analysis was performed by SRL, with inputs from
SM, HSM, MMS, and ICA. All authors led the writing of the manuscript and contributed critically to the
drafts and gave final approval for publication.

Funding This study was supported in part by the Biological Dynamics of Forest Fragments Project
(BDFFP—INPA), through the Thomas Lovejoy research fellowship program 2014/2015.

References
Adams JM, Kang C, June-Wells M (2014) Are tropical butterflies more colorful? Ecol Res 29:685–691
Arima EY, Barreto P, Araújo E, Soares-Filho B (2014) Public policies can reduce tropical deforestation:
lessons and challenges from Brazil. Land Use Policy 41:465–473. https://doi.org/10.1007/s11284-014-
1154-1
Barlow J, Lennox GD, Ferreira J et al (2016) Anthropogenic disturbance in tropical forests can double
biodiversity loss from deforestation. Nature 535:144–147
Bates HW (1862) Contributions to an insect fauna of the Amazon valley (Lepidoptera: Heliconidae). Trans
Linn Soc Lond 23:495–566
Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat
Softw 67:1–48
Bierregaard RO, Lovejoy TE, Kapos V, Santos AA, Hutchings RW (1992) The biological dynamics of
tropical rainforest fragments: a prospective comparisonof fragments and continuous forest. Bioscience
11:859866
Briolat ES, Burdfield-steel ER, Paul SC, Katja HR, Seymoure BM, Stankowich T, Stuckert AMM (2018)
Diversity in warning coloration: selective paradox or the norm? Biol Rev 94:388–414
Brown KS (2005) Geologic, evolutionary, and ecological bases of the diversification of neotropical but-
terflies: implications for conservation. In: Dick CW, Moritz G (eds) Tropical rainforest: past, present,
and future. The University of Chicago Press, Chicago, pp 166–201
Brown KS Jr, Hutchings RW (1997) Disturbance, fragmentation, and the dynamics of diversity in Ama-
zonian forest butterflies. In: Laurance WF, Bierregaard RO (eds) Tropical forest remnants: ecology,
management and conservation of fragmented communities, 1st edn. The University of Chicago Press,
Chicago, pp 91–110
Bruno JF, Cardinale BJ (2008) Cascading effects of predator richness. Front Ecol Environ 6:539–546

123
 Biodiversity and Conservation

Caro T, Stoddard MC, Stuart-fox D (2017) Animal coloration research: why it matters. Philos Trans R Soc B
372:20160333
Chao A, Wang YT, Josy L (2013) Entropy and the species accumulation curve: a novel entropy estimator
via discovery of new species. Methods Ecol Ecol 4:1091–1100
Checa MF, Donoso DA, Rodriguez J, Levy E, Warren A, Willmott K (2018) Combining sampling tech-
niques aids monitoring of tropical butterflies. Insect Conserv Divers 12:362–372
Cheng W, Xing S, Chen Y, Lin R, Bonebrake TC, Nakamura A (2018) Dark butterflies camouflaged from
predation in dark tropical forest understories. Ecol Entomol 43:304–309
Ciuti S, Northrup JM, Muhly TB, Simi S, Musiani M, Pitt JA, Boyce MS (2012) Effects of humans on
behaviour of wildlife exceed those of natural predators in a landscape of fear. PLoS ONE 7:e50611
Crouzeilles R, Ferreira MS, Chazdon RL, Lindenmayer DB, Sansevero JBB, Monteiro L, Iribarrem A,
Latawiec AE, Strassburg BBN (2017) Ecological restoration success is higher for natural regeneration
than for active restoration in tropical forests. Sci Adv 3 (11):e1701345
Cuthill IC, Bennett ATD, Partridge JC, Maier EJ (1999) Plumage reflectance and the objective assessment
of avian sexual dichromatism. Am Nat 153:183–200
Cuthill IC, Allen WL, Arbuckle K et al (2017) The biology of color. Science 80:357
Dalrymple RL, Kemp DJ, Flores-Moreno H, Laffan SW, White TE, Hemmings FA, Tindall ML, Moles AT
(2015) Birds, butterflies and flowers in the tropics are not more colourful than those at higher latitudes.
Glob Ecol Biogeogr 24:1424–1432
Dalrymple RL, Flores-Moreno H, Kemp DJ, White TE, Laffan SW, Hemmings FA, Hitchcock TD, Moles
AT (2018) Abiotic and biotic predictors of macroecological patterns in bird and butterfly coloration.
Ecol Monogr 88:204–224
Delhey K, Peters A (2016) Implications for conservation of anthropogenic impacts on visual communication
and camouflage. Conserv Biol 31:1–36
Ellers J, Boogs CL (2004) Functional ecological implications of intraspecific differences in wing
melanization in Colias butterflies. Biol J Linn Soc 82:79–87
Endler JA (1993) The color of light in forests and its implications. Ecol Monogr 63:1–27
Endler JA, Mielke PW (2005) Comparing entire colour patterns as birds see them. Biol J Linn Soc
86:405–431
Endler JA, Mappes J (2017) The current and future state of animal coloration research. Phil Trans Royal Soc
B: Biol Sci 372 (1724):20160352
Freitas AVL, Iserhard CA, Santos JP, Yasmin J, Carreira O, Ribeiro DB, Henrique D, Melo A, Henrique A,
Rosa B, Marini-filho OJ, Accacio GM (2014) Studies with butterfly bait traps: an overview. Rev
Colomb Entomol 40:203–212
Hook TV (1997) Insect coloration and implications for conservation. Fla Entomol 80:193–210
Hsieh TC, Ma KH, Chao A (2016) iNEXT: an R package for rarefaction and extrapolation of species
diversity (Hill numbers). Methods Ecol Ecol 7:1451–1455
Instituto Nacional de Pesquisas Espaciais—INPE (2019) Projeto Prodes—Monitoramento da floresta
Amazônica brasileira por satélite. https://www.obt.inpe.br/prodes/. Accessed 18 March 2019
Jakovac CC, Peña-Claros M, Kuyper TW, Bongers F (2015) Loss of secondary-forest resilience by land-use
intensification in the Amazon. J Ecol 103:67–77
Johnson H, Solensky MJ, Satterfield DA, Davis AK (2014) Does skipping a meal matter to a butterfly’s
appearance? Effects of larval food stress on wing morphology and color in monarch butterflies. PLoS
ONE 9:1–9
Joshi J, Prakash A, Kunte K (2017) Evolutionary assembly of communities in butterfly mimicry rings. Am
Nat 189:58–76
Kingsolver JG (1985) Thermoregulatory significance of wing melanization in Pieris butterflies (Lepidoptera:
Pieridae): physics, posture, and pattern. Oecologia 66:546–553
Komdeur JM, Oorebeek M, Van Overveld T, Cuthill IC (2005) Mutual ornamentation, age, and reproductive
performance in the European starling. Behav Ecol 16:805–817
Laurance WF, Vasconcelos HL (2009) Consequências ecológicas da fragmentação florestal na amazônia.
Oecologia Bras 13:434–451
Maia R, Eliason CM, Bitton PP, Doucet SM, Shawkey MD (2013) Pavo: An R package for the analysis,
visualization and organization of spectral data. Methods Ecol Evol 4:906–913
Mappes J, Kokko H, Ojala K, Lindström L (2014) Seasonal changes in predator community switch the
direction of selection for prey defences. Nat Commun 5:5016
Mesquita RDCG, Massoca PEDS, Jakovac CC, Bentos TV, Williamson GB (2015) Amazon rainforest
succession: stochasticity or land-use legacy? Bioscience 65:849–861
Metzger JP, Bustamante MMC, Ferreira J et al (2019) Why Brazil needs its legal reserves. Perspect Ecol
Conserv. https://doi.org/10.1016/j.pecon.2019.07.002

123
Biodiversity and Conservation

Papageorgis C (1975) Mimicry in Neotropical butterflies: why are there so many different wing-coloration
complexes in one place? Am Sci 63:522–532
Pellissier L, Albouy C, Bascompte J, Farwig N, Graham C, Loreau M, Maglianesi MA, Melián CJ, Pitteloud
C, Roslin T, Rohr R, Saavedra S, Thuiller W, Woodward G, Zimmermann NE, Gravel D (2018)
Comparing species interaction networks along environmental gradients. Biol Rev 93:785–800
Pinheiro CEG, Freitas AVL, Campos VC, DeVries PJ, Penz CM (2016) Both palatable and unpalat-
able butterflies use bright colors to signal difficulty of capture to predators. Neotrop Entomol
45:107–113
R Core Team (2018) R: a language and environment for statistical computing. R Foundation for Statistical
Computing, Vienna
Roslin T, Andrew NR, Asmus A et al (2017) Higher predation risk for insect prey at low latitudes and
elevations. Science 356:742–744
Rozendaal DMA, Bongers F, Aide TM (2019) Biodiversity recovery of Neotropical secondary forests. Sci
Adv 5:eaau3114
Schtickzelle N, Joiris A, Van Dyck H, Baguette M (2007) Quantitative analysis of changes in movement
behaviour within and outside habitat in a specialist butterfly. BMC Evol Biol 7:1–15
Soga M, Kawahara T, Fukuyama K, Sayama K, Kato T, Shimomura M, Itoh T, Yoshida T, Ozaki K (2015)
Landscape versus local factors shaping butterfly communities in fragmented landscapes: does host
plant diversity matter? J Insect Conserv 19:781–790
Spaniol RL, Duarte LDS, Mendonça MDS Jr, Iserhard CA (2019) Combining functional traits and phy-
logeny to disentangling Amazonian butterfly assemblages on anthropogenic gradients. Ecosphere
10:e02837
Spottiswoode CN, Stevens M (2012) Host-parasite arms races and rapid changes in bird egg appearance. Am
Nat 179:633–648
Stevens M (2005) The role of eyespots as anti-predator mechanisms, principally demonstrated in the
Lepidoptera. Biol Rev 80:573–588
Stevens M (2016) Cheats and deceits: how animals and plants exploit and mislead. Oxford University Press,
Oxford
Stevens M, Ruxton GD (2014) Do animal eyespots really mimic eyes? Curr Zool 60:26–36
Stevens M, Parraga CA, Cuthill IC, Partridge JC, Troscianko TS (2007) Using digital photography to study
animal coloration. Biol J Linn Soc 90:211–237
Stevens M, Stoddard MC, Higham JP (2009) Studying primate color: towards visual system dependent
methods. Int J Primatol 30:893–917
Stevens M, Lown AE, Wood LE (2014) Colour change and camouflage in juvenileshore crabs Carcinus
maenas. Front Ecol Evol 2:14
Stork NE (2010) Re-assessing current extinction rates. Biodivers Conserv 19:357–371
Talloen W, Van Dyck H, Lens L (2004) The cost of melanization: Butterfly wing coloration under envi-
ronmental stress. Evolution 58:360–366
Troscianko J, Stevens M (2015) Image calibration and analysis toolbox—a free software suite for objec-
tively measuring reflectance, colour and pattern. Methods Ecol Evol 6:1320–1331
Vieira RRS, Pressey RL, Loyola R (2019) The residual nature of protected areas in Brazil. Biol Conserv
233:152–161
Walton OC, Stevens M (2018) Avian vision models and field experiments determine the survival value of
peppered moth camouflage. Commun Biol 1:118
Willmott KR, Willmott JRC, Elias M, Jiggins CD (2017) Maintaining mimicry diversity: optimal warning
colour patterns differ among microhabitats in Amazonian clearwing butterflies. Proc R Soc B
284:20170744

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

Affiliations

Ricardo Luı́s Spaniol1 • Milton de Souza Mendonça Jr.1 •

Sandra Maria Hartz1 • Cristiano Agra Iserhard2 • Martin Stevens3

123
 Biodiversity and Conservation

& Ricardo Luı́s Spaniol

[email protected]
1
Programa de Pós-graduação Em Ecologia, Universidade Federal Do Rio Grande Do Sul,
Porto Alegre, Rio Grande do Sul, Brazil
2
Programa de Pós-graduação Em Biologia Animal, Universidade Federal de Pelotas,
Capão Do Leão, Rio Grande do Sul, Brazil
3
Centre for Ecology & Conservation, College of Life & Environmental Sciences, University of
Exeter, Penryn Campus, Penryn, Cornwall TR10 9FE, UK

123