Journal of Insect Conservation

, Volume 21, Issue 3, pp 545–557 | Cite as

Butterfly communities respond to structural changes in forest restorations and regeneration in lowland Atlantic Forest, Paraná, Brazil

  • John ShueyEmail author
  • Paul Labus
  • Eduardo Carneiro
  • Fernando Maia Silva Dias
  • Luis Anderson R. Leite
  • Olaf H. H. Mielke


The Atlantic forest is one of the most diverse biomes on Earth but human activities are transforming this ecosystem into one of the most endangered. Most remnant old-growth rainforest is embedded within a mosaic of regenerating forest, tree plantations, pastures, and agricultural production. This has left a large percentage of the region’s endemic species threatened with extinction. Butterflies are considered as sensitive indicators of ecological conditions, especially in the Atlantic forest. This community can provide a window into animal response to restoration and how recovering habitats are used by native animal communities. The primary goal of this paper was to determine if butterfly communities respond to measures of structural recovery in naturally regenerating and re-forested pastures, and if this response increases the similarity of recovering butterfly communities relative to those of intact forests. Butterfly communities were sampled using two sampling methodologies, passive bait trapping and timed meander counts. These data sets were combined and correlated to assessment of habitat structure. We found that butterfly communities respond rapidly to structural changes in habitats as forest structure recovers on abandoned and restored pastures. While many species of mature forest inhabiting butterflies use regenerating forests as habitat, our young forests also retained an almost intact community of ruderal pasture inhabiting butterflies as well, indicating that these habitats retain many features of highly disturbed pastures. We suggest that measures of beta-diversity, which can be used to assess convergence in community structure, are far superior to the alpha-diversity measures that are typically used for assessing restoration recovery.


Beta-diversity Biodiversity Ecology Ecological recovery Lepidoptera Papilionoidea 


The Atlantic Forest is among the most diverse and most threatened ecosystems of the world. Originally covering over 1.3 million km2, and extending for more than 3000 km along the eastern Brazilian coast, this forest is considered a biodiversity hotspot and is a priority area for conservation due to its high degree of endemism (Myers et al. 2000). Unfortunately, the Atlantic Forest is heavily impacted by human activities and just 11–16% of the region remains forested (Ribeiro et al. 2009) and just 7% of this is primary forest (Dean 1996). The majority of the remaining remnant forest is embedded in a mosaic of regenerating forest, tree plantations, agricultural production and human development (Ribeiro et al. 2009). This has left a large percentage of the region’s endemic species threatened with extinction.

In contrast to the general trend of extreme deforestation in the Atlantic Forest, in the state of Paraná in southern Brazil large remnants of forests still exist, especially along the mountainous range of Serra do Mar, forming an ecological mosaic composed of swaths of mature forest interspersed with secondary and regenerating forest in various stages of succession. Many of these regenerating forests are in abandoned or replanted water buffalo pastures. There are still many active buffalo pastures scattered throughout the lowlands in the region and these pastures are increasingly viewed as potential restoration sites to reduce threats within conservation areas (Ferretti and de Britez 2006). While the functional diversity of such forests have been questioned relative to undisturbed forests (e.g., Bihn et al. 2010), there is little doubt that regenerating forest improves ecological integrity and function relative to the alternative agricultural use, which is often dominated by near monocultures of non-native, invasive African grasses (such as Brachiaria spp.). In fact, successful conservation of Atlantic Forest in more highly degraded regions of Brazil will be largely dependent upon the successful restoration and regeneration of secondary tropical forest to enlarge and re-connect scattered remnants of primary forest.

In Paraná, forest recovery in abandoned or reforested pastures is fairly rapid and predictable. Abandoned pastures are quickly colonized by pioneer species of trees that are animal-dispersed, and by 8 years post-abandonment, more conservative species begin to establish (Cheung et al. 2010; Siminski et al. 2011; Sobanski and Marques 2014). Forest structure changes rapidly with an increase in tree abundance, stem volume and species richness as canopy closure increases (Guariguata and Ostertag 2001; Marques et al. 2014). Herbaceous recovery is less well documented, but personal observations indicate that invasive African pasture grasses decrease in cover as the young forest canopy becomes more closed and that native species establish in the more shaded environment. These successional processes proceed even more rapidly when fast growing native pioneer trees are planted in the pastures as the first step towards restoration (Ferretti and de Britez 2006; Sobanski and Marques 2014).

This research was initiated to examine the initial response of butterfly communities to reforestation of pasture land. We chose a species rich assemblage of phytophagous insects, butterflies (Lepidoptera), that provide a potential regional fauna of between 500 and 700 species for our analysis (Mielke 1994; Brown et al. 2000; Dolibaina et al. 2011; Francini et al. 2011). Butterflies are increasingly viewed as sensitive indicators of ecological conditions, including the Atlantic Forest (Barlow et al. 2007a; Brown et al. 2000; Freitas et al. 2014). Butterfly communities can provide a window into animal response to restoration and how recovering habitats are used by native animal communities (Samways 1997). In this regard we specifically assessed butterfly communities across a diversity of sites that represent high-quality lowland forests, their local degradation endpoints (buffalo pastures) and regenerating and reforested pastures using sampling techniques that provide temporal snapshots of butterfly community structure. Our primary goal was to determine how butterfly communities respond to structural/ecological recovery in restored and regenerating pastures, and if this response increases the similarity of recovering butterfly communities relative to those of intact forests. Our data also allow us to calculate measures of both α-diversity and β-diversity in response to ecological healing, and we used these to compare the relative utility of these as measures of ecological recovery. Finally, our methods were designed to assess sampling effort and sample size on community similarly analyses using measures of β-diversity.


This research was conducted in the coastal plain area of the state of Paraná, in the municipalities of Antonina and Guaraqueçaba (Fig. 1). Originally the region was covered by contiguous ombrophilous lowland and submontane tropical forests, but these forests suffered massive exploitation and in the lowlands were largely converted to buffalo pastures (Ferretti and Britez 2006). Today’s landscape is a mosaic of open land, secondary forests and scattered, but relatively large patches of old-growth forests.

Fig. 1

Research sites on coastal plain of Paraná. The research plots were located in the municipalities of Antonina and Guaraqueçaba. Site codes for the map: RF1-3 Remnant Forest plots 1 though 3; P1-2 pasture plots 1 and 2; ReF1-2 reforestation plots 1 and 2, and ; NR natural regeneration forest plot

Our sample sites were chosen to represent points across the local ecological and disturbance gradients that define natural and anthropogenic lowland forest communities in the region including; intact lowland forest; young naturally regenerating forest in old pastures; tree plantings in old pastures; and open pastures. In order to minimize differences between potential habitat types, our sample sites were located exclusively on level, poorly drained lowland forests and former forests at elevations between 8 and 20 m. To eliminate the effects of dispersal distances, pastures, reforestation and regeneration forests plots sampled were all adjacent to mature forest habitats. We set a minimum habitat plot size of 40 ha to help ensure that we sampled resident communities. Insect communities in habitat patches smaller than this are likely to be highly influenced by edge effects. There were pre-existing trails in place though all the plots prior to the proposed work and we used these trails for sampling. Replicates of specific habitat types were located at least 1 km from one another. These sites are located in two private nature reserves ‘‘Reserva Natural do Rio Cachoeira’’ and ‘‘Reserva Natural Serra do Itaqui’’ owned and managed by the Brazilian NGO ‘‘Society for wildlife research and environmental education’’ (SPVS). These private reserves are part of the greater Environmental Protection Area of Guaraqueçaba. Our study areas are part of an ecological restoration program intended to restore connectivity and continuity across the lowland forest (Feretti and Britez 2006) and to reduce historic forest fragmentation due to conversion to buffalo pastures. We sampled eight sites, representing four habitat types.

Remnant forest

Three sample sites characterized by mature trees, an intact canopy broken only by scattered tree-fall gaps, and a dense shrub layer. These sites have all experienced some level of selective tree harvesting in the past.

Natural regeneration

One sample site that was an old pasture surrounded by remnant forest. This site was removed from grazing 14 years prior to sampling, allowing natural tree regeneration to proceed from adjacent forests. Other than widely scattered “pasture trees”, trees are young and canopy cover is patchy. Understory shrubs are patchy but occasionally dense. Tall pasture grasses dominate the ground layer under canopy openings and along access roads.


Two sites in old pastures that were planted to mixes of 15–20 early successional tree species 12 and 14 year prior to sampling. Both sites are adjacent to remnant forests. Other than widely scattered “pasture trees”, trees are young and canopy cover is patchy. Understory shrubs are patchy but occasionally dense. Tall pasture grasses dominate the ground layer under canopy openings and along access roads used to maintain the tree plantings.


Two actively grazed sample sites surrounded by remnant forest. Other than widely scattered pasture trees, these habitats are wide open, dominated by short, gazed pasture grasses.

Butterfly communities were sampled using two sampling methodologies in order to maximize our coverage of the butterfly assemblages present at each site—passive bait trapping and timed meander counts. We sampled each site three times, once each in January, March and April, 2011. Bait trap samples were used to assess the rich fauna that is attracted to fermenting fruit (Freitas et al. 2014). Four standard butterfly traps were placed 1–2 m above ground at 50 m intervals near the center of the sample plots. Traps were baited using fermenting bananas as described in Shuey (1997). For each sampling event, traps were operated for three consecutive days (72 h) and serviced daily to replenish baits and to record the number and species of butterflies sampled. Timed butterfly meander counts were conducted in each plot to assess species not attracted to bait. Two replicate 1-h sample periods were recorded for each sampling event. Counts were conducted during periods of clear weather and between the hours of 10:00 am and 3:00 pm to ensure consistency of the samples. During the sample periods, every butterfly encountered was either identified in flight or captured and vouchered for future identification. Some unrecognized species escaped vouchering and were not included in the data set. Data were combined from the two sampling methods to create a snapshot of butterfly communities at each site during each sampling period. Vouchers are held in the Shuey collection and will ultimately be deposited in the entomology collections at the U.S. Natural History Museum in Washington D.C.

It was beyond our capacity to assess plant communities using traditional quantitative species-based assessments in a forest renowned for its botanical diversity (Murray-Smith et al. 2008). Instead we measured structural components of the habitat which intuitively differentiate between our three habitat types. Our goal was to align our habitat types on a structural gradient between mature remnant forest and highly disturbed pastures that corresponds to the recognized trend of forest recovery in the area (Cheung et al. 2010).

We established a 200 m transect at each site to assess these three structural variables;

  1. 1.

    at 10 m points we used a Haglof Vertex Hypsometer II to measure canopy height. Tree height increases with forest age;

  2. 2.

    at these same 10 m points we used a densiometer to measure of percent canopy opening. Percent canopy opening decreases with forest age and there is almost no canopy in pastures;

  3. 3.

    we visually estimated percent cover of Brachiaria spp. to the nearest 10% at these points. Percent cover of Brachiaria spp. increases with tree removal and is absent in remnant forest.


These measures were intended to place our reforested and regenerating plots on a successional continuum between species poor pastures, and species rich remnant forest based on conceptual model of Atlantic forest succession presented by Cheung et al. (2010).

Data were examined to provide insights into potential patterns of community diversity and relationships among the sites. Following Pielou (1975), we calculated Brillouin D indices for the quantitative butterfly samples as a measure of the diversity of a collection using routines in Species Diversity and Richness 4 (Seaby and Henderson 2007). Behavior differences between species influence detection probability between species and Brillouin D is best used in situations where a collection is made, but sampling was non-random or the full composition of the community is known (Pielou 1975). Because our sampling effort, although standardized by time and effort, was potentially inconsistent between sites (based on the number of individual butterflies recorded), we also calculated both predicted number of species based on statistically equalized sampling effort (rarefaction) and Fisher’s Alpha for each site (using Seaby and Henderson 2007).

After data normalization (required for canopy and Brachiaria cover), we used multiple regression (Minitab Release 14) to determine if our habitat structure measures predicted the number of butterfly species recorded at each site, the predicted number of species based on rarefaction analysis, and Brillouin D indices. We produced cluster analysis dendrograms to illustrate relationships among sample sites using Community Analysis Package version 4.0 Henderson and Seaby (2007). Habitat structure variables were clustered using average linkage and distance measures to construct dendrograms. Butterfly data were clustered separately using additive similarity trees based on Bray–Curtis community coefficients of similarity based on species presence–absence and as well as abundance data for butterflies to determine how sample sites related to one another. We used Ward’s clustering protocols to join groups. At each iteration all possible pairs of groups are compared and the two groups chosen for fusion are those which will produce a group with the lowest variance. A SIMPROF test (Clarke et al. 2008) was therefore used to check whether the presence of each cluster group is different from others that could be obtained by chance. This analysis adds a statistical meaning to ecological similarity clusters by providing a significant value upon each group of samples, whose null hypothesis regards the lack of community structure.

We used Mantel tests to test for significance of congruence in community similarity between all pair-wise comparisons between habitat structure and butterfly and communities. The Mantel test is a statistical test of the correlation between two similarity matrices. Mantel tests were performed using PopTools software (Hood 2006) with 999 random permutations.

To determine if our data set was sensitive to sampling size and effort, we also tested for significance of congruence for each sampling event (N = 3), all pairwise combinations of sampling events (N = 2) and of course the entire data set. We used R2 values from the Mantel analyses to determine if increased sampling effort increased the strength of relationships between habitat structure matrices and butterfly community matrices as measured by increased R2 values.


Plant and butterfly communities

The structural characteristics of the plant communities varied in predictable ways (Table 1). Average tree height was greatest in remnant forest, and lowest in pastures (p < 0.05). Canopy opening was greatest in pasture and lowest in remnant forest (p < 0.05). The percent cover of Brachiaria was highly correlated with canopy opening (p < 0.05, R2 = 55.5%) and was greatest in pastures.

Table 1

Summary of habitat structure and butterfly communities


Remnant forest

Natural regeneration


Pasture 1










Plant community structure

 Avg tree height









 Avg canopy opening









 Avg % Brachiaria cover (%)









Butterfly community

 Number of species









 Est. num. species (rarefaction)









 Fisher’s alpha









 Brillouin D









All forest structural measures differ significantly by general habitat type (p < 0.05 for remnant forest, reforested/regeneration and pasture). Number of species per habitat type and diversity measures (Brillouin D) do not significantly differ between general habitat types (p = 0.62 and p = 0.35 respectively)

A total of 3547 butterflies representing 213 species were encountered during this study. Eighty-seven species were recorded from single sites. Only one measure of α-diversity (Table 1) produced a significant trend across our ecological gradient. The estimated number of species based on rarefaction was significantly different between sites (p = 0.04): remnant forest (\(\bar{\text{x}}\) =  69.4 ± 11.3) sites supported the most species followed by regeneration/restoration (\(\bar{\text{x}}\) =  56.9 ± 11.3) and pastures (\(\bar{\text{x}}\) =  35.4 ± 2.5): by this measure remnant forest supports significantly more species than do pastures. Although not significant (p = 0.34) Brillouin D indices followed a similar pattern and were highest in remnant forest (\(\bar{\text{x}}\) = 3.155 ± 0.40) followed by regeneration and restoration sites (\(\bar{\text{x}}\) = 3.14 ± 0.34), and pastures (  \(\bar{\text{x}}\) = 2.70 ± 0.11). In contrast, both actual counts of species, and species richness as indicated by Fisher’s alpha indicate potential but non-significant reordering of α-diversity measures. The number of species encountered during sampling (p = 0.62) was highest in regeneration/plots (\(\bar{\text{x}}\)  = 72.7.3 ± 27.7) followed by remnant forest (\(\bar{\text{x}}\)  = 70.0 ± 11.5) and pastures (\(\bar{\text{x}}\)  = 5.5 ± 5.0). Likewise, species richness estimates based on Fisher’s alpha (p = 0.13), and were highest in regeneration/restoration sites (\(\bar{\text{x}}\)  = 32.4 ± 8.8) followed by remnant forest sites (\(\bar{\text{x}}\) = 24.6 ± 9.0) and pastures (\(\bar{\text{x}}\)  = 4.0±0.7).

Based on multiple regression, there were no relationships between the structural measures of habitat measures and α-diversity. There were no consistent ecological trends from the number of species observed by site (p = 0.71, R2 = 27%), the estimated number of species based on rarefaction (p = 0.07, R2 = 79%), Fisher’s Alpha (p = 0.30, R2 = 56%) and Brillouin D (p = 0.51, R2 = 41%). These results are consistent with other previous research demonstrating that such simple measures of Lepidoptera communities seldom track neatly with habitat assessments or ecological gradients (Howard et al. 1998; Oertli et al. 2005; Fiedler and Truxa 2012; Shuey et al. 2012).

Table 2 presents qualitative species richness from all habitat types and combinations of habitats. Although difficult to interpret, more species were encountered in forested sites. Perhaps more enlightening, 66 species were found exclusively in remnant forest sites, double the species that were encountered exclusively in the reforestation + regeneration sites (33 species) or pastures (18 species).

Table 2

Summary of species richness by habitat types and habitat type combinations, including total number of species for each category, and number of species found only in general habitat types and combinations


Remnant forest

Remnant + restoration


Restoration + pasture


Remnant + pasture

All habitat types

Total # of butterfly species encountered in each habitat or habitat combination








Total # of species restricted to each habitat type or habitat combination








A total of 213 butterfly species were recorded

Community similarity analyses and congruence

Dendrograms derived from habitat structure measures and butterfly communities are illustrated in Figs. 2 and 3. The dendrograms for habitat structure and butterfly communities identify three significant clustered assemblages, each relating to the primary habitat types sampled—remnant forest, pasture, and regenerating forest. Moreover, habitat structure relationships demonstrate that for the characteristics we assessed, that regenerating forest is more similar to remnant forest than to open pasture. There is nothing surprising about the clustering of habitats based on structure measures. We choose to assess easily measured variables that we felt a priori, would provide insight into the structural differences between the three key habitat types assessed, remnant forest, regenerating forest and pasture. Conceptually, we understand that restored and regenerating forests are intermediate in these characters, and the results confirm this.

Fig. 2

Dendrograms of community relationships based on similarity ordination. a Plant community structural measures; b Butterfly by species abundances; c Butterfly by species presence only

Fig. 3

Dendrograms of butterfly community relationships using SIMPROF to determine statistical significance of each cluster of samples. Thin lines indicate similarity between samples in clusters is not different from that expected by chance, while thick lines indicate statistically significant clusters. Left Butterfly by species abundance; Right Butterfly by species presence only

In contrast, both cluster analyses of butterflies (abundance and presence only) produced dendrograms with a dichotomy between remnant forests and disturbed habitats (both pastures and regenerating forests). Restored and regenerating forest butterfly communities, although quite distinct, are more similar to pastures than to intact forest. Community similarity between habitat structure and butterfly communities (using both presence only and species abundance) was significantly positively correlated (p < 0.001, R2 = 0.41 and 0.25 respectively) (Fig. 4). Habitats that were structurally similar also supported statistically distinct and similar butterfly communities.

Fig. 4

Community correlations between habitat structure and butterflies. Top forest structure versus butterfly species abundance. Bottom forest structure versus butterfly species presence only. Both correlations are significant (p < 0.05)

Visual analysis of the data underlying these relationships is revealing (Tables 3, 4). For species found at two or more sample sites, 20 were restricted exclusively to remnant forest, seven were limited to restored/regenerating forest and three were exclusive to pastures. Perhaps more interestingly, 35 species were found in both remnant forest and restored/regenerating forests but not in pastures. This clearly indicates that a subset of the butterfly community that requires forest habitat finds the recovering forest to be acceptable habitat. Similarly, 30 species were found in pastures and reforested/regenerating forests, but not in remnant forests, which strongly influence clustering patterns. This indicated that recovering forest habitats remain very disturbed habitats and support virtually all ruderal butterfly species that characterize pastures in the region. Thus, while several forest butterflies are colonizing these recovering forest habitats, almost all pasture inhabiting species are still present in the restorations.

Table 3

Ecological distribution of species associated with forested habitats and found at two or more sample sites (representing a subset of species tallied in Table 2 which includes species found at single sample sites as well)

The majority of species fall into discernable ecological groupings that underlie the structure of dendrograms in Figs. 1 and 2 (boxes outlined with solid lines highlight these groupings)

Impact of sampling effort on community similarly analyses

In every possible permutation of sample combinations, our butterfly community similarity matrices were significantly correlated to habitat structure matrices (all Mantel P’s < 0.001). However, when we correlated R2 values against sampling effort, there are indications of a positive relationship (Fig. 5), but no significant increases in predictive power for either the data from the quantitative set or presence/absence data (p = 0.2 and p = 0.8 respectively). For this data set, increasing sampling effort did not increase community resolution as measured by increased R2 values. Thus, while each discrete sample event creates a statistically significant snapshot of butterfly community structure, increased sampling does not significantly increase the resolution of the data. We assume that this is the result of our relatively small sample size relative to the total species pool and ecological noise present in the system. For the quantitative samples, we suspect that the positive trend apparent in Fig. 5 would continue with increased sampling effort and eventually become statistically significant. We have no such expectations for data sets reduced to species presence/absence.

Fig. 5

The relationship between sampling effort and predictive value of butterfly community similarity matrices when correlated to habitat similarity matrices. While there are indications that R2 increases with increased sample size, the trend is not significant for either the data from the quantitative set or presence absence (p = 0.2 and p = 0.8 respectively)


Assessing the success or failure of ecological restoration is difficult and most typically involves assessments of abiotic conditions and/or plant communities relative to reference conditions or establishment of plant materials (Ruiz-Jaen and Aide 2005; Cristescu et al. 2013). Typical measures include α-diversity, vegetation structure and ecological processes. While restoration normally succeeds well in establishing plant communities, these newly established plant communities typically differ significantly in composition and structure from reference sites. Such results are generally interpreted within the context of early successional dynamics and observed differences are attributed to successional pathways towards full recovery (Suding 2011; Hansen and Gibson 2014). These interpretations may well be valid, but avoid answering the question at hand at our study sites—do restorations function ecologically as better “forest habitat” than they were functioning prior to restoration?

Increasingly, animal communities are being used to assess restoration impacts and ecological function (eg, Bihn et al. 2010; Grimbacher and Catterall 2007; Sáfián et al. 2011; Ribeiro et al. 2010). Because they are fairly easy to monitor and can represent various functional and trophic levels, invertebrates are often used. Most of these studies employ a traditional evaluation of restored communities versus the desired reference condition. Using measures of α-diversity such as measures of calculated species richness or various diversity indices, success is generally interpreted as increased “richness and diversity” approaching that of the target community, regardless of actual species composition, and these measures have generated inconsistent and often contradictory results (e.g., Kati 2004; Barlow et al. 2007b). For example, Howard et al. (1998) reported little congruence in species richness between plants, Lepidoptera and vertebrates in Ugandan forests. Likewise, in the Swiss Alps, Oertli et al. (2005) report no relationship between bees, grasshoppers, and aculeate wasps. Shuey et al. (2012) found no relationship between the α-diversity of moth species and plant species in North American oak barrens, sand prairies and their degradation endpoints. In assessing butterfly communities across a forested disturbance gradient in Ghana, Sáfián et al. (2011) reported that estimated (Chao1) butterfly species richness in clear-cut and farmland habitats adjacent to forests was high relative to recovering and undisturbed forest habitats. They interpreted their results as indicative of the nature of temporary gap habitats in rainforests, which are capable of supporting mobile species of open habitats, but which are also habitat for forest species that utilize and disperse through open habitats in low numbers.

Thus, it should come as no surprise that our measures of α-diversity across our sites create conflicting (and generally not statistically significant) trends. Similar to Sáfián et al. (2011) and Ribeiro et al. (2010), some of our measures, such as observed butterfly species richness and Brillouin D, trend higher in reforestations/restorations. But this is not because the communities are “restored” but rather because these habitats retain virtually the entire ruderal community present prior to restoration, while adding a significant portion of the targeted forest community in response to structural changes associated with forest recovery (see Tables 2, 3, 4). Given that we find no significant correlations between our structural measures of habitat and any measure of α-diversity, we do not believe that simple measures of α-diversity reflect ecological recovery in this system.

Table 4

Ecological distribution of species associated with disturbed habitats and found at two or more sample sites (representing a subset of species tallied in Table 2 which includes species found at single sample sites as well)

The majority of species fall into discernable ecological groupings that underlie the structure of dendrograms in Figs. 1 and 2 (boxes outlined with solid lines highlight these groupings). A dotted line highlights anomalous association of species found in both highly disturbed pastures and mature forest

Despite our inconclusive results generated by measures of α-diversity, our analysis of the same data sets indicate that measures of β-diversity are highly and significantly congruent with our simple assessment of the successional gradient Figs. 2,3 and 4. Although we sampled butterfly communities three times for this effort, each sample, when analyzed seperately, produces essentially the same result and high level of significance. While there are indeed indications of increased data resolution as measured by R2 values with increasing sampling effort (Fig. 5), relevant insight into restoration trajectory can be obtained by analyzing a single data set that captures the essence of community structure [see Filgueiras et al. (2016) for a discussion of this for the fruit-feeding guild in Brazil]. Moreover, we believe that repeated sampling over ecologically relevant time spans may capture changing relationships among sample sites as ecological restoration gradients heal, providing insight into ecological functionality of restored habitats (Sant’Anna et al. 2014).

In support of biodiversity conservation, restorations strive to support recognizable biotic communities—not “community facsimiles” as measured by species counts, diversity indices or some functional assessment of community structure (e.g., Bihn et al. 2010; Howard et al. 1998). Real biological communities are predictable assemblages of species, and this predictability is the underlying factor that must be assessed in restorations that target such endpoints. In other words, does the restored community have “authenticity” relative to the desired (reference) ecological condition? Measures of β-diversity are more relevant measures for conservation and ecological functioning, and we suggest that convergence patterns of β-diversity are more fitting assessments of ecological recovery endpoints. Studies that assess community similarity across definable ecological gradients within similar communities typically find signs of congruence among taxonomic assemblages across habitat gradients (Grimbacher and Catterall 2007, Su et al. 2004; Shuey et al. 2012; Carneiro et al. 2014). Here we used measures of community similarity on the assumption that a successful restoration trajectory intended to heal forested ecosystems would be indicated by the convergence toward our restoration goal (remnant forest) and divergence of restored sites away from the pre-restoration state (pasture), similar to the approach used by Grimbacher and Catterall (2007) and Sáfián et al. (2011). These studies from Australia and Africa, both demonstrated that as forest structure recovers, they are quickly colonized by many insect species that are typically restricted to rainforest habitats. In both cases, these individual species responses assembled into communities that began to resemble authentic forest insect communities.

Our results from the Atlantic Rainforest corroborate this ecological convergence and are indicative of ecological communities that are in flux, but pointed along a positive trajectory. We demonstrate that butterfly communities are unambiguously responsive to changes in plant community structure along the habitat gradient we sampled. As structural relationships change between the sample sites, there are corresponding and consistent changes in the butterfly community as forest dwelling species populate restored habitats. But these restorations still support an essentially intact community of ruderal species found in pastures—over 90% of the species found in pasture habitats are also found in restoration plots. This implies that many of the plant species that support ruderal butterflies found in pasture settings are still present in the recovering forests we sampled. This is certainly the case for Brachiaria, the dominant forage grass that we assessed across all habitat types (and a likely hostplant for many of the Hesperiidae we found in pastures and restorations). We assume, that as these young forests mature, that Brachiaria and other weedy species from pastures will continue to decline producing a concurrent decline in ruderal species of butterflies.



Staff at the Sociedade de Pesquisa em Vida Selvagem e Educação Ambiental welcomed and facilitated our field work and were gracious hosts during our prolonged visits. This work was made possible by Leandro Baumgarten, Science Manager for The Nature Conservancy’s program in Brazil.


  1. Barlow J, Overal W, Araujo I, Gardner T, Peres C (2007a) The value of primary, secondary and plantation forests for fruit-feeding butterflies in the Brazilian Amazon. J Appl Ecol 44:1001–1012CrossRefGoogle Scholar
  2. Barlow J, Gardner T, Araujo I, Avila-Pires T, Bonaldo A, Costa J, Esposito M, Ferreira L, Hawes J, Hernandez M, Hoogmoed M, Leite R, Lo-Man-Hung N, Malcolm J, Martins M, Mestre L, Miranda-Santos R, Nunes-Gutjahr A, Overal W, Parry L, Peters S, Ribeiro-Junior M, da Silva M, Motta C, Peres C (2007b) Quantifying the biodiversity value of tropical primary, secondary, and plantation forests. Proc Natl Acad Sci 104:18555–18560CrossRefPubMedPubMedCentralGoogle Scholar
  3. Bihn J, Gebauer G, Brandl R (2010) Loss of functional diversity of ant assemblages in secondary tropical forests. Ecology 91:782–792CrossRefPubMedGoogle Scholar
  4. Brown J, Brown K, Freitas A (2000) Atlantic forest butterflies: indicators for land-scape conservation. Biotropica 32:934–956CrossRefGoogle Scholar
  5. Carneiro E, Mielke O, Casagrande M, Fiedler K (2014) Community structure of skipper butterflies (Lepidoptera, Hesperiidae) along elevational gradients in Brazilian Atlantic Forest reflects vegetation type rather than altitude. PLoS ONE 9:e108207CrossRefPubMedPubMedCentralGoogle Scholar
  6. Cheung K, Liebsch D, Marques M (2010) Forest recovery in newly abandoned pastures in Southern Brazil: implications for the Atlantic Rain Forest Resilience. Nat Conserv 8:66–70CrossRefGoogle Scholar
  7. Clarke K, Somerfield P, Gorley R (2008) Testing of null hypotheses in exploratory community analyses: similarity profiles and biota-environment linkage. J Exp Mar Biol Ecol 366:56–69CrossRefGoogle Scholar
  8. Cristescu R, Rhodes J, Frere C, Banks P (2013) Is restoring flora the same as restoring fauna? Lessons learned from koalas and mining rehabilitation. J Appl Ecol 50:421–431CrossRefGoogle Scholar
  9. Dean W (1996) A ferro e fogo—a história e a devastação da Mata Atlântica Brasileira. Companhia das Letras, São PauloGoogle Scholar
  10. Dolibaina D, Mielke O, Casagrande M (2011) Borboletas (Papilionoidea e Hesperioidea) de Guarapuava e arredores, Paraná, Brasil: um inventário com base em 63 anos de registros Butterflies (Papilionoidea and Hesperioidea) from Guarapuava and vicinity, Paraná, Brazil: an inventory based on records of 63 years. Biota Neotropica 11:341–354CrossRefGoogle Scholar
  11. Ferretti A, de Britez R (2006) Ecological restoration, carbon sequestration and biodiversity conservation: the experience of the Society for Wildlife Research and Environmental Education (SPVS) in the Atlantic Rain Forest of Southern Brazil. J Nat Conserv 14:249–259CrossRefGoogle Scholar
  12. Fiedler K, Truxa C (2012) Species richness measures fail in resolving diversity patterns of speciose forest moth assemblages. Biodivers Conserv 21:2499–2508CrossRefGoogle Scholar
  13. Filgueiras B, Melo D, Leal I, Tabarelli M, Freitas A, Iannuzzi L (2016) Fruit-feeding butterflies in edge-dominated habitats: community structure, species persistence and cascade effect. J Insect Conserv 20:539–548CrossRefGoogle Scholar
  14. Francini R, Duarte M, O Mielke, Caldas A, Freitas A (2011) Butterflies (Lepidoptera, Papilionoidea and Hesperioidea) of the “Baixada Santista” region, coastal São Paulo, southeastern Brazil. Rev Brasil Entomol 55:55–68CrossRefGoogle Scholar
  15. Freitas A, Iserhard C, dos Santos J, Carreira J, Ribeiro D, Melo D, Rosa A, Filho O, Accacio G, Uehara-Prado M (2014) Studies with butterfly bait traps: an overview. Rev Colomb Entomol 40:209–218Google Scholar
  16. Grimbacher P, Catterall C (2007) How much do site age, habitat structure and spatial isolation influence the restoration of rain forest beetle species assemblages? Biol Conserv 135:107–118CrossRefGoogle Scholar
  17. Guariguata M, Ostertag R (2001) Neotropical secondary forest succession: changes in structural and functional characteristics. For Ecol Manag 148:185–206CrossRefGoogle Scholar
  18. Hansen M, Gibson D (2014) Use of multiple criteria in an ecological assessment of a prairie restoration chronosequence. Appl Veg Sci 17:63–73CrossRefGoogle Scholar
  19. Henderson P, Seaby R (2007) Community analysis package version 4.0, Pisces Conservation Ltd, LymingtonGoogle Scholar
  20. Hood G (2006) PopTools version 2.7.5.
  21. Howard P, Viskanic P, Davenport T, Kigenyi F, Baltzer M, Dickenson C, Lwanga J, Matthews R, Balmford A (1998) Complementarity and the use of indicator groups for reserve selection in Uganda. Nature 392:472–475Google Scholar
  22. Kati V, Devillers P, Dufrênec M, Legakisd A, Vokoue D, Lebrunf P (2004) Hotspots, complementarity or representativeness? designing optimal small-scale reserves for biodiversity conservation. Biol Conserv 120:471–480CrossRefGoogle Scholar
  23. Marques C, Zwiener V, Ramos F, Borgo M, Marques R (2014) Forest structure and species composition along a successional gradient of Lowland Atlantic Forest in Southern Brazil. Biota Neotropica 14:1–11Google Scholar
  24. Mielke C (1994) Papilionoidea e Hesperioidea (Lepidoptera) de Curitiba e seus arredores, Paraná, Brasil, com notas taxonômicas sobre Hesperiidae. Rev Brasil Zool 11:759–776CrossRefGoogle Scholar
  25. Murray-Smith C, Brummitt N, Oliveira-Filho A, Bachman S, Moat J, Lughadha E, Lucas D (2008) Plant diversity hotspots in the Atlantic Coastal Forests of Brazil. Conserv Biol 23:151–163CrossRefPubMedGoogle Scholar
  26. Myers N, Mittermeier R, Mittermeier C, Fonseca G, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403:853–858CrossRefPubMedGoogle Scholar
  27. Oertli S, Müller A, Steiner D, Breitenstein A, Dorn S (2005) Cross-taxon congruence of species diversity and community similarity among three insect taxa in a mosaic landscape. Biol Conserv 126:195–205CrossRefGoogle Scholar
  28. Pielou E (1975) Ecological diversity. Wiley, New YorkGoogle Scholar
  29. Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM (2009) The Brazilian Atlantic Forest: how much is left, and how is the remaining forest distributed? Implications for conservation. Biol Conserv 142:1141–1153CrossRefGoogle Scholar
  30. Ribeiro DB, Prado PI, Brown KSJR, Freitas AVL (2010) Temporal diversity patterns and phenology in fruit-feeding butterflies in the Atlantic forest. Biotropica 42:710–716CrossRefGoogle Scholar
  31. Ruiz-Jaen M, Aide T (2005) Restoration success: how is it being measured? Restor Ecol 13:569–577CrossRefGoogle Scholar
  32. Sáfián S, Csontos G, Winkler D (2011) Butterfly community recovery in degraded rainforest habitats in the Upper Guinean forest zone (Kakum forest, Ghana). J Insect Conserv 15:351–359CrossRefGoogle Scholar
  33. Samways M (1997) Insect conservation biology. Chapman and Hall, New YorkGoogle Scholar
  34. Sant’Anna C, Ribeiro D, Garcia L, Freitas A (2014) Fruit-feeding butterfly communities are influenced by restoration age in tropical forests. Restor Ecol 22:480–485CrossRefGoogle Scholar
  35. Seaby R, Henderson P (2007) Species diversity and richness 4, Pisces Conservation Ltd, LymingtonGoogle Scholar
  36. Shuey J (1997) An optimized portable bait trap for quantitative sampling of butterflies. Trop Lepid 8:1–4Google Scholar
  37. Shuey J, Metzler E, Tungesvick K (2012) Moth communities correspond with plant communities in Midwestern (Indiana, USA) sand prairies and oak barrens and their degradation endpoints. Am Midl Nat 167:273–284CrossRefGoogle Scholar
  38. Siminski A, Fantini A, Guries R, Ruschel A, dos Reis M (2011) Secondary forest succession in the Mata Atlantica, Brazil: floristic and phytosociological trends. ISRN Ecology 2011 Article ID 759893, p 19Google Scholar
  39. Sobanski N, Marques M (2014) Effects of soil characteristics and exotic grass cover on the forest restoration of the Atlantic Forest region. J Nat Conserv 22:217–222CrossRefGoogle Scholar
  40. Su J, Debinski C, Jakubauskas M, Kindscher K (2004) Beyond species richness: community similarity as a measure of cross-taxon congruence for coarse-filer conservation. Conserv Biol 18:167–173CrossRefGoogle Scholar
  41. Suding K (2011) Toward an era of restoration in ecology: successes, failures, and opportunities ahead. Ann Rev Ecol Evol Syst 42:465–487CrossRefGoogle Scholar

Copyright information

© Springer International Publishing Switzerland 2017

Authors and Affiliations

  1. 1.The Nature ConservancyIndianapolisUSA
  2. 2.The Nature ConservancyMerrillvilleUSA
  3. 3.Laboratorio de Estudos em Lepidoptera Neotropical, Departamento de ZoologiaUniversidade Federal do ParanáCuritibaBrazil

Personalised recommendations