Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Clade‐specific consequences of climate change to amphibians in Atlantic Forest protected areas

Clade‐specific consequences of climate change to amphibians in Atlantic Forest protected areas Climate change is a major threat to biodiversity ( Root et al. 2003 , Araujo et al. 2004 , Pounds et al. 2006 , Leadley et al. 2010 ), since it induces species’ range shifts and alters patterns of species diversity and distribution, including species occurrence inside protected areas (PAs; ( Araujo et al. 2004 , Araújo et al. 2011 , Monzón et al. 2011 , Loyola et al. 2012 )). This is critical given that PAs are the cornerstone of conservation strategies and 10–15% of the Earth's surface is currently under some form of legal protection ( Rodrigues et al. 2004 , Chape et al. 2005 ). If these places fail to represent species in the future, then current biodiversity crisis would reach unprecedented levels. The disappearance of amphibians from within PAs is one of the most puzzling aspects of the worldwide decline of amphibians, demanding conservation plans for this group ( Stuart et al. 2004 , Brito 2008 , Lips et al. 2008a , b, Loyola et al. 2008 ). Habitat alteration, exotic species introductions, diseases, UV‐B radiation, and climate warming have been implicated in the decline of some species ( Araújo et al. 2006 , Lips et al. 2008a ). According to the International Union for the Conservation of Nature ( IUCN 2012 ), Neotropical amphibians have the higher proportion of threatened species, as well as several records of population declines. In particular, amphibians seem to be a group severely affected by climate change ( Pounds et al. 2006 , Lips et al. 2008b , Nori et al. 2011 ). These high levels of population declines and species threat are creating demands for effective strategies for conservation in order to mitigate the impacts of climate change ( Hof et al. 2011 , Lemes and Loyola 2013 , Loyola et al. 2013 ). However, most studies aiming at the conservation of amphibians are focused on species, ignoring other key aspects of biodiversity. Scientists are now challenged to develop predictive models capable of summarizing important ecological and evolutionary processes while producing robust recommendations for conservation actions ( Sutherland et al. 2006 , Bielby et al. 2009 ). Yet, when other biodiversity aspects are considered in conservation studies, they are primarily expressed by means of phylogenetic and functional diversity ( Faith 1992 , Carvalho et al. 2010 , Safi et al. 2011 ). Phylogenetic diversity adds the species evolutionary relatedness into the diversity measures ( Faith 1992 , Hardy and Senterre 2007 ). Phylogenetic diversity is an important biodiversity component to be conserved, given that it represents the evolutionary history of conservation target groups ( Faith 1992 , Sechrest et al. 2002 , Trindade‐Filho et al. 2012 ). Nevertheless, since all measures of phylogenetic diversity aim to synthesize phylogenetic information, other dimensions of biodiversity end up being neglected. Two sites could have similar phylogenetic diversity levels, while harboring completely different phylogenetic lineages ( Duarte 2011 ). Further, extinction risk is not independent of evolutionary history ( Purvis et al. 2000 , Corey 2010 ), since closely‐related species often show similar ecological requirements ( Harvey and Pagel 1991 , Wiens et al. 2010 ) and, consequently, similar susceptibility to extinction. These facts suggest that phylogenetic composition, in particular, may be a key driver of threatened species distribution at broad spatial scales. Here, we evaluate current and future climatic suitability of PAs located in the Atlantic Forest Biodiversity Hotspot, Brazil, and its implications for amphibian conservation. More specifically, we addressed the following questions: 1) how will climate change affect the geographical pattern of amphibian species richness, diversity, and phylogenetic diversity in the region? 2) How does phylogenetic composition of amphibian assemblages respond to climate change? 3) What clades are being favored or hindered with changes in climate? We focused our analyses on the Atlantic Forest Biodiversity Hotspot, which holds 18% of all South American amphibian species, many of them endemics ( Carnaval et al. 2009 ). Further, today only ca 9% of the remaining forest and 1% of the original forests are legally protected ( Ribeiro et al. 2009 ). An effective reserve network is therefore imperative to address conservation investment in appropriate sites taking into account climate change to ensure the species and evolutionary history persistence in the long term. Methods Species data We rasterized digital range maps of 431 amphibians species inhabiting the Atlantic Forest obtained from IUCN database (< www.iucnredlist.org/technical‐documents/spatial‐data#amphibians >) to an 10 × 10 km resolution grid, totaling 11 461 grid cells. As a representation of species distribution, range maps are necessarily a scale‐dependent abstraction ( Hurlbert and White 2005 , Hurlbert and Jetz 2007 ). In general, these maps are typically drawn a smoothed outline around records of occurrence interpolated and can include commission errors or false presences ( Hurlbert and White 2005 ). It should be an initial approach to seed species distributions models and to identify general priorities, which can be refined as more accurate information is obtained for regions with limited information about presence and absence of species ( Lemes et al. 2011 , Rondinini et al. 2011 ). Nevertheless, these maps have been used for modeling species’ ecological niches at broad geographic scales ( Diniz‐Filho et al. 2009 , Lawler et al. 2009 ). We did not use point location data for modeling species’ ecological niches. Although commonly used, point location data are sparse and often biased in their sampling toward areas that are easily accessible, thus increasing omission errors. For these reasons, here we used digital range maps to generate a presence–absence matrix of amphibian occurrence in the Atlantic Forest (see also Lemes and Loyola 2013 ). This matrix, along with climate variables, was then used as our input data for building species’ ecological niche models. Ecological niche models (ENMs) We obtained current climatic data from the WorldClim database (< www.worldclim.org/current >) and future climatic scenarios from CIAT ( ) through WorldClim website. The IPCC's Fourth Assessment Report (AR4) developed these future scenarios. We modeled the ecological niche of each species as a function of four climatic variables: annual mean temperature, temperature seasonality (standard deviation × 100), annual precipitation, and precipitation seasonality (coefficient of variation). These variables are often used to explain geographical/macroecological patterns in amphibian species richness ( Wiens et al. 2006 , Buckley and Jetz 2007 ). Variables for future climate came from three coupled Atmosphere‐Ocean General Circulation Models (AOGCMs), for the A2 scenario for the year 2080: CCCMA‐CGCM2, CSIRO‐MK2 and HCCPR‐HadCM3. The selected AOGCMs are widely used in the literature, having also different equilibrium climate sensitivity values ranging from 3.1°C to 4.4°C. Equilibrium climate sensibility is the annual mean surface air temperature change experienced by the climate system after it has attained a new equilibrium in response to a doubling of CO 2 concentration, and are within the range of all AOGCMs available from IPCC ( IPCC 2007 ). There are several methods for modeling species’ ecological niche and projecting their distributions ( Franklin 2009 ) and there is no consensus on the best method to predict the presence and absence of species according to environmental variables. To cope with these problems, a combination of different projections built upon different conditions and methods – the ensemble forecasting approach ( Araújo and New 2007 ) – proved to be more interpretative than single model analysis. This approach minimizes methodological uncertainties given that the final solution is a unique consensus weighted by the overall statistical fit of combined models, allowing also to quantify and map uncertainties associated with ENMs ( Diniz‐Filho et al. 2009 ). We chose six ENMs: generalized linear models – GLM ( McCullagh and Nelder 1989 ), generalized additive models – GAM ( Yee and Mitchell 1991 ), multivariate adaptive regression splines – MARS ( Muñoz and Felicísimo 2004 ), genetic algorithm for rule set production – GARP ( Stockwell and Noble 1992 ), maximum entropy – MaxEnt ( Phillips et al. 2006 ), and random forest – RF ( Breiman 2001 ). We used both presence‐only and presence–absence methods for modeling species’ ecological niches. Our models can be grouped in two types of methods ( Pearson 2007 ). Methods that use ‘background’ climatic data for the whole study area (e.g. MaxEnt). These methods evaluate how the climatic conditions where species are known to occur relate to the climate across the rest of the study area (the ‘background’); and methods that generate (sample) ‘pseudo‐absences’ from the study area (e.g. GARP). These methods assess differences between the occurrence sites and a set of sites chosen from the study area, which are used instead of real absence data. Here, the set of ‘pseudo‐absences’ were selected randomly ( Stockwell and Noble 1992 ). We evaluated ENMs predictive ability by partitioning our data into training and test sets (at the ratio of 75:25, respectively), with ten replicates of random partitions. To examine the species richness inside PAs and other sites, we established a threshold of pseudo‐absences for each model to allow building the receiving operating curve (ROC) and transform quantitative predictions of models into a binary vector of 0/1, indicating forecasted presences or absences in each grid cell outside species’ extent of occurrences ( Allouche et al. 2006 ). We established the cut‐off point by using the delimitation of the bioclimatic envelope of 95%, allowing us to generate the ROC. Finally, we adopted the True Skill Statistics (TSS) as our measure of model fit. Sensitivity and specificity were calculated based on the probability threshold to which their sum is maximized, not being affected by prevalence. TSS values range from −1 to + 1, where + 1 indicates a perfect fit, minimizing overprediction and omission error rates; values close to −1 indicate a performance worse than those randomly expected ( Allouche et al. 2006 ). We did the ensembles of forecasts to produce more robust predictions and reduce the uncertainties owing to the modeling. We projected distributions to future climate and obtained 180 projections per species within each set of methods (6 modeling methods × 3 AOGCMs × 10 randomly partitioned data) and 60 projections per species for the present (6 modeling methods × 10 randomly partitioned data) – this allowed us to generate a frequency of projections in the ensemble. We then generated the frequency of projections weighted by the TSS statistics for each species and timeframe ( Fig. 1 ). We considered the presence of a species only in cells with 50% or more of frequency of projections, but we held a continuous value when this occurred. This procedure produced two consensus maps, one for the present and one for the future, which were used as input data to analyze amphibian occurrence inside PAs as well as the phylogenetic diversity and composition. 1 Maps of modeled amphibian species richness based on geographic range overlap of 431 species predicted to occur in the Atlantic Forest Hotspot for (A) present and (B) future climatic conditions. Protected areas are shown in black. Amphibian occurrence in PAs We obtained the location of PAs (IUCN categories I–IV) in the Atlantic Forest from the World Database on Protected Areas ( ). We overlaid PA polygons onto our grid considering a grid cell as ‘protected’ even if only a small part of it holds a PA; we assumed therefore that all species occurring in that cell could potentially benefit from the occurrence of a PA in that cell. The incidence of amphibian species in each PA polygon was obtained from consensus maps generated by ENMs based on present or future climatic conditions. Thus, each PA polygon was described by the presence of amphibian species in either present or future climatic conditions. We computed the number of species and Simpson diversity index for each PA polygon under present and future climatic conditions, which was used as a measure of species richness and diversity, respectively. To evaluate the effect of climate change on the number of amphibian species and Simpson diversity, we used paired t‐tests, comparing diversity patterns in the future against present. By doing so, we guaranteed that each PA was compared in terms of present and future amphibian diversity patterns only with itself. Phylogenetic diversity and composition We described each PA polygon by the number of amphibian species per genus for phylogenetic analyses. The phylogenetic tree of amphibian genera from the Atlantic Forest was a pruned version from Pyron and Wiens (2011) . This tree is uses 12 loci from 2871 species. These loci are broadly sampled and have successfully been used in amphibian systematics. A phylogenetic patristic distance matrix ( D F ) for all genera included in Pyron and Wiens’ phylogenetic hypothesis was computed using the software Mesquite 2.73 ( Maddison and Maddison 2010 ) after arbitrary ultrametrization. Posteriorly, we pruned from D F all those genera not occurring in PA polygons, yielding 57 amphibian genera. We also removed seven genera occurring in PA polygons but not included in Pyron and Wiens's tree. We computed phylogenetic diversity using Rao quadratic entropy ( Hardy and Senterre 2007 , Pavoine and Bonsall 2011 ) for each PA polygon under either present or future climatic conditions. To evaluate the effect of climate change on phylogenetic diversity, we used paired t‐tests, comparing diversity patterns in the future against present. We also evaluated phylogenetic composition of amphibian genera using the phylogenetic fuzzy‐weighting method ( Pillar and Duarte 2010 ) and implemented in the R packages SYNCSA ( Debastiani and Pillar 2012 ) and ape ( Paradis et al. 2004 ). This metric uses phylogenetic similarities between taxa to scale‐up the phylogenetic relationships from taxa to the site level. Pairwise phylogenetic distances in D P were transformed into a phylogenetic similarity matrix ( S P = 1 − D P ). Then, phylogenetic similarities in S P were used to weight amphibian genera composition in each PA polygon under present or future climatic conditions, using a fuzzy set algorithm ( Pillar and Duarte 2010 ). This procedure generated a matrix P containing the phylogeny‐weighted genera composition for each PA polygon, under present or future climatic conditions. That is, the presence of genera i in a given PA polygon was shared with each genera j occurring in the whole set of PA polygons (considering both present and future climatic conditions), taking into account the phylogenetic similarity between i and j . Accordingly, those genera j more closely‐related to i (e.g. from the same family) received a proportionally higher fraction of the presence of i in that PA polygon than more phylogenetically distant genera (e.g. from a different order), which will receive a proportionally lower fraction, and so on. Therefore, matrix P expresses the clade composition in PA polygon under present or future climatic conditions. We did a principal coordinates analysis ( Gower 1966 ) on matrix P using square‐rooted Bray–Curtis dissimilarities between PA polygons as resemblance measure to generate principal coordinates of phylogenetic structure (PCPS) for amphibian assemblages. Each PCPS is an eigenvector describing an independent phylogenetic gradient in the dataset ( Duarte 2011 ). The PCPS with the highest eigenvalue describes broader phylogenetic gradients related to deeper tree nodes across protected areas, such as that connecting anurans and caecilians. As the eigenvalues of the other PCPS decrease, finer phylogenetic gradients related to higher nodes (e.g. families, genera) are described ( Duarte et al. 2012 ). PCPS analysis generated 367 eigenvectors with positive eigenvalues. The first four PCPS extracted from matrix P explained > 57% of the total variation in phylogenetic composition in amphibian assemblages ( Table 1 ). 1 Effect of climate change on phylogenetic composition of amphibian genera occurring in protected areas in the Atlantic Forest Biodiversity Hotspot, Brazil. Variable Eigenvalue Contribution to trace (%) t 186 p‐value PCPS 1 2.28 33.64 −31.11 < 0.001 PCPS 2 0.75 11.02 1.82 0.071 PCPS 3 0.54 8.02 −1.30 0.194 PCPS 4 0.31 4.64 1.26 0.209 Trace 6.77 100.00 Each PCPS was analyzed separately using paired t‐tests to evaluate the effect of climate change on phylogenetic composition. PCoA was run in MULTIV 2.63b statistical software (by V. Pillar, available at , user's guide included). Paired t‐tests were performed using SigmaPlot 11 (Systat Software). Results For most species, TSS values were relatively high (TSS ± SD = 0.63 ± 1.33) indicating good model fit. Combined model projections indicated areas of high species richness in central and eastern parts of the Atlantic Forest in the present ( Fig. 1A ). This region should decrease in species richness by 2080, whereas northern and southwestern portions should face many species extinction ( Fig. 1B ). All modeling methods forecasted a general reduction in species’ ranges, leading to a decrease in the number of sites with high species richness. Most species had a significant range contraction (up to 85%) and 12% of species were expected to be regionally extinct (Supplementary material Appendix 1, Table A1). Not surprisingly, climate change had a negative impact on both species richness ( Fig. 2A ) and Simpson diversity ( Fig. 2B ). Nonetheless, phylogenetic diversity increased under future climatic conditions, albeit such increase was not clear ( Fig. 2C ). 2 (A) Number of species, (B) species diversity (Simpson index), and (C) phylogenetic diversity (Rao quadratic entropy) of amphibians modeled according to present and future climatic conditions in Atlantic Forest protected areas. Only PCPS 1 (33% of total variation) showed significant association with climate change, while PCPS 2 (11%) showed marginal significance ( Table 1 ). The occurrence of amphibian genera in PAs under future climatic conditions was negatively related to PCPS 1, while amphibian distribution under present climate was positively related to PCPS 1. Amphibian composition along PCPS 2 in PAs under present climate was negatively associated with this ordination axis, whereas amphibian distribution in the future was positively related to PCPS 2 ( Table 1 ). The ordination scatterplot shows that the occurrence of Gymnophiona, Hemiphractidae, and Pipidae in PAs was positively related to climate change, whereas the occurrence of late‐divergent clades, such as Ceratophryidae, Craugas toridae, Cycloramphidae, Centrolenidae, Eleu therodactylidae, Hylodidae, Hylidae, Microhylidae, and Strabomantidae was negatively associated with climate change in PAs ( Fig. 3 ). Not surprisingly, the distribution of Lithobates , an invasive species in the Atlantic Forest, tended to be positively related to PCPS 1, indicating that climate change will likely favor that species. 3 Scatter diagram of the first two principal coordinates of phylogenetic structure (PCPS) of amphibian genera occurring in Atlantic Forest protected areas. Genera occurrence was modeled according to present (empty squares) and future (grey squares) climatic conditions. Black dots represent species grouped in monophyletic clades in the biplot. Discussion We forecasted shifts and contractions in species’ range for most Atlantic Forest amphibians. On the one hand, most PAs would become climatically unsuitable to sustain their current number of species under climate change, leading to a decrease in species richness and diversity. On the other hand, phylogenetic diversity tended to increase given the higher representativeness of basal clades (Gymnophiona and Pipidae) under future climatic conditions. That is to say that species‐specific and clade‐specific responses to climate change could be largely different, and suggests that amphibian responses to climate change shows a low degree of phylogenetic signal. This is an important message in this paper. This result contradicts the phylogenetic patterns found by Thuiller et al. (2011) . In that study, the authors observed that under a climate change scenario in Europe, mammal, plant, and bird assemblages tended to loose phylogenetic diversity, suggesting that sensitivity to climate change is to some extent phylogeneticaly conserved. Further, our results also provided clues for analyzing the pattern of phylogenetic diversity in a more refined way than provided by other similar methods ( Peres‐Neto et al. 2012 ), given that we were able to visualize how each clade responds to climate change. Families negatively associated with climate change generally have more specialized reproductive modes ( Haddad and Prado 2005 ), such as direct development, with egg laying in humid forest floors (Craugastoridae, Eleutherodactylidae, Strabomantidae, and some Micro hylidae genera), or breed only in primary forest streams (Centrolenidae, Cycloramphidae, Hylodidae). Species of those lineages have narrow habitat requirements for reproduction and foraging, which could make them more prone to extinction or range contractions, likely due to decreased forest cover as a consequence of high temperatures. On the other hand, Pipidae and Gymnophiona tended to increase their representativeness in PAs along the Atlantic Forest. The current distribution of the aquatic genus Pipa in the Atlantic Forest is restricted to warm regions in the northeastern ( Duellman 1999 , Frost 2012 ). With climate change and range shifts, it could expand its distribution southwards, increasing overall phylogenetic diversity of PAs. Other families seem not to be largely affected by climate change, such as Bufonidae and Leptodactylidae. Species of these families are generalists and occur in both open and forested habitats (e.g. Rhinella ) or have reproductive modes independent of water or humid places (e.g. Leptodactylus ; Haddad and Prado 2005 ). Future climate change has also potential implications for the spread of diseases, such as the deadly fungus Batrachochytrium dendrobatidis (Bd) ( Hof et al. 2011 ). It is known that the optimum temperature for Bd growth and development ranges from 17°C to 25°C, with temperatures higher than 30°C being lethal ( Blaustein et al. 2012 ). Therefore, some lineages may benefit from predicted increasing temperatures, since the immune system of amphibians may become more efficient. However, factors affecting susceptibility of species and lineages to infection are still controversial. For example, several factors with a strong phylogenetic pattern seem to play a key role in the resistance to Bd infection ( Blaustein et al. 2012 ), such as peptides and bacteria from amphibian's skin ( Maciel et al. 2003 ), body size, and reproductive mode ( Bancroft et al. 2011 ). Conversely, other studies ( Crawford et al. 2010 ) suggest that a phylogenetic component is not involved in epidemic events of Bd infection in Central America. The relationship between current and future climates and diseases‐driven declines is complex ( Blaustein and Kiesecker 2002 ) and still demands research efforts. Further, our findings indicate that the most likely impact of climatic changes should occur along the coastal line of the Atlantic Forest. This results concur with Lemes and Loyola (2013) and add up to finding of Hof et al. (2011) . These authors found a synergetic effect of climate change, land‐use change and disease spreading in this region. In their study, however, effects of climate change were more pronounced in the north of Atlantic Forest. Menéndez‐Guerrero and Graham (2013) also found synergetic effects of climate change and chytridiomicosis exert great impact on amphibians in Ecuador. These results indicate that climate change might foster infection by pathogens and that management of local populations and disease control are essential for safeguarding amphibians in this regions, in which Bd has been already found ( Eterovick et al. 2005 , Verdade et al. 2012 ). In any case, however, there should be no major changes in species richness due to short latitudinal temperature gradients of the tropics ( Colwell et al. 2008 ). Consequently range shifts will be likely more pronounced toward higher elevations than higher latitudes ( Bush and Hooghiemstra 2005 ). Indeed, our projections for the future predict many species ranges moving upward in elevation, which concentrates a large number of endemic species in the Atlantic Forest ( Carnaval et al. 2009 ), such as brachycephalids that occur in isolated mountaintops (e.g. Brachycephalus pernix ; Pombal Jr et al. 1998 ), and several hylid genera (e.g. Bokermannohyla, Phasmahyla ). In fact, climatically suitable sites tend to concentrate in the mountain range of Serra do Mar, in the eastern part of the biome – exactly where we forecast a high species richness in the future. The whole region is filled with forest refugia from the Pleistocene ( Porto et al. 2012 ) and that might explain why climate is not changing drastically in these portions of the biome. Such reduction of climatically suitable sites leads to an overall range contraction that defines the association of clades to present or future climatic conditions. Clades showing lower range contraction (up to 25%; e.g. Gymnophiona, Pipidae, Supplementary material Appendix 1, Table A1) relate to future climatic conditions. Those with a high range contraction (> 40%; e.g. Aromobatidae, Dendrobatidae, Supplementary material Appendix 1, Table A1) are underrepresented in the future, being related to current climatic conditions. Thus, the increase in phylogenetic diversity is not necessarily good news. It means that, once many clades are loosing climate space in the future, and at the same time, basal clades are being favored, phylogenetic diversity increases in PAs, while species richness decreases. All in all, we are loosing species representation in PAs, while evolutionary information is being maximized by the maintenance of phylogenetically distant clades. Recent studies found similar results on the likely range contraction of the American bullfrog Lithobates catesbeianus ( Giovanelli et al. 2008 , Nori et al. 2011 , Loyola et al. 2012 ), one of the hundred worst invasive species in the world ( Lever 2003 ). As highlighted by Loyola et al. (2012) , a retraction in suitable climatic conditions for L. catesbeianus in western Brazil could drive the alien species into Atlantic Forest PAs. Other studies agree that the bullfrog is expected to increase its representation along Atlantic Forest PAs ( Giovanelli et al. 2008 , Nori et al. 2011 ). Such spread of the species in PAs explains why Lithobates is related to future climatic conditions of PAs in our study. The occurrence of the bullfrog in Atlantic Forest PAs is especially concerning, given that it has deleterious effects on populations of native amphibians and other organisms through competition and predation ( Giovanelli et al. 2008 , Nori et al. 2011 ). As any study based on species distribution models, our analyses have their own caveats. Predictions of range shifts for amphibian species in the Atlantic Forest are clearly dependent of two assumptions: unlimited dispersal ability and absence of biological interactions ( Soberón 2007 ). Model predictions are fraught with uncertainties to obtain range changes ( Diniz‐Filho et al. 2009 ), beyond the stemming ecological and evolutionary lousily represented in the models ( Guisan and Thuiller 2005 , Brook et al. 2009 ). The ensemble forecasting is an alternative and conservative approach for reducing uncertainties among models, which combines the central tendency of multiples bioclimatic niche models ( Araújo and New 2007 ). Consensual models, when appropriately analyzed, may recover species range shifts in climate change context at a broad scale ( Marmion et al. 2009 , Dobrowski et al. 2011 ). Nevertheless, given the uncertain nature of model predictions, our results should be taken with prudence. Further, our models did not include dispersal and adaptation components of species’ distribution. Norberg et al. (2012) , for example, developed an eco‐evolutionary model of multi‐species responses to climate change. They showed that dispersal and adaptation mediate extinction risks in different ways, leading to biodiversity change through time and across changing climates. Integrating these aspects into ecological niche modeling is still in its infancy and new studies in this field are need to allow for applying it for multiple species and broad geographic scales. Incorporating the phylogenetic structure of amphibian assemblages into our analyses show that we should expect a clade‐specific effect of climate change on the distribution of amphibians inhabiting Atlantic Forest PAs. Basal clades should be less affected, while clades retaining specialized reproductive modes should be highly impacted. Identifying major changes in the phylogenetic pool represent a first step towards a better understanding of how assembly processes related to climate change will affect ecological communities. We hope our results contribute to a more deep analysis of the impacts of climate change not only on species, but also on the evolutionary relationships at the local and regional assemblages. Acknowledgements RDL and LDSD received a productivity scholarship awarded by the CNPq. PL is supported by a CNPq PhD scholarship. FTB and DBP receives a CAPES PhD scholarship. The research was funded by the FTC (Brazil/Portugal) program of CAPES, by the Brazilian Network for the study of Climate Change (MCT/Rede CLIMA), and by Conservation International, Brazil. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecography Wiley

Clade‐specific consequences of climate change to amphibians in Atlantic Forest protected areas

Loading next page...
 
/lp/wiley/clade-specific-consequences-of-climate-change-to-amphibians-in-v62ixLjXBg

References (52)

Publisher
Wiley
Copyright
"Copyright © 2014 Wiley Subscription Services, Inc., A Wiley Company"
ISSN
0906-7590
eISSN
1600-0587
DOI
10.1111/j.1600-0587.2013.00396.x
Publisher site
See Article on Publisher Site

Abstract

Climate change is a major threat to biodiversity ( Root et al. 2003 , Araujo et al. 2004 , Pounds et al. 2006 , Leadley et al. 2010 ), since it induces species’ range shifts and alters patterns of species diversity and distribution, including species occurrence inside protected areas (PAs; ( Araujo et al. 2004 , Araújo et al. 2011 , Monzón et al. 2011 , Loyola et al. 2012 )). This is critical given that PAs are the cornerstone of conservation strategies and 10–15% of the Earth's surface is currently under some form of legal protection ( Rodrigues et al. 2004 , Chape et al. 2005 ). If these places fail to represent species in the future, then current biodiversity crisis would reach unprecedented levels. The disappearance of amphibians from within PAs is one of the most puzzling aspects of the worldwide decline of amphibians, demanding conservation plans for this group ( Stuart et al. 2004 , Brito 2008 , Lips et al. 2008a , b, Loyola et al. 2008 ). Habitat alteration, exotic species introductions, diseases, UV‐B radiation, and climate warming have been implicated in the decline of some species ( Araújo et al. 2006 , Lips et al. 2008a ). According to the International Union for the Conservation of Nature ( IUCN 2012 ), Neotropical amphibians have the higher proportion of threatened species, as well as several records of population declines. In particular, amphibians seem to be a group severely affected by climate change ( Pounds et al. 2006 , Lips et al. 2008b , Nori et al. 2011 ). These high levels of population declines and species threat are creating demands for effective strategies for conservation in order to mitigate the impacts of climate change ( Hof et al. 2011 , Lemes and Loyola 2013 , Loyola et al. 2013 ). However, most studies aiming at the conservation of amphibians are focused on species, ignoring other key aspects of biodiversity. Scientists are now challenged to develop predictive models capable of summarizing important ecological and evolutionary processes while producing robust recommendations for conservation actions ( Sutherland et al. 2006 , Bielby et al. 2009 ). Yet, when other biodiversity aspects are considered in conservation studies, they are primarily expressed by means of phylogenetic and functional diversity ( Faith 1992 , Carvalho et al. 2010 , Safi et al. 2011 ). Phylogenetic diversity adds the species evolutionary relatedness into the diversity measures ( Faith 1992 , Hardy and Senterre 2007 ). Phylogenetic diversity is an important biodiversity component to be conserved, given that it represents the evolutionary history of conservation target groups ( Faith 1992 , Sechrest et al. 2002 , Trindade‐Filho et al. 2012 ). Nevertheless, since all measures of phylogenetic diversity aim to synthesize phylogenetic information, other dimensions of biodiversity end up being neglected. Two sites could have similar phylogenetic diversity levels, while harboring completely different phylogenetic lineages ( Duarte 2011 ). Further, extinction risk is not independent of evolutionary history ( Purvis et al. 2000 , Corey 2010 ), since closely‐related species often show similar ecological requirements ( Harvey and Pagel 1991 , Wiens et al. 2010 ) and, consequently, similar susceptibility to extinction. These facts suggest that phylogenetic composition, in particular, may be a key driver of threatened species distribution at broad spatial scales. Here, we evaluate current and future climatic suitability of PAs located in the Atlantic Forest Biodiversity Hotspot, Brazil, and its implications for amphibian conservation. More specifically, we addressed the following questions: 1) how will climate change affect the geographical pattern of amphibian species richness, diversity, and phylogenetic diversity in the region? 2) How does phylogenetic composition of amphibian assemblages respond to climate change? 3) What clades are being favored or hindered with changes in climate? We focused our analyses on the Atlantic Forest Biodiversity Hotspot, which holds 18% of all South American amphibian species, many of them endemics ( Carnaval et al. 2009 ). Further, today only ca 9% of the remaining forest and 1% of the original forests are legally protected ( Ribeiro et al. 2009 ). An effective reserve network is therefore imperative to address conservation investment in appropriate sites taking into account climate change to ensure the species and evolutionary history persistence in the long term. Methods Species data We rasterized digital range maps of 431 amphibians species inhabiting the Atlantic Forest obtained from IUCN database (< www.iucnredlist.org/technical‐documents/spatial‐data#amphibians >) to an 10 × 10 km resolution grid, totaling 11 461 grid cells. As a representation of species distribution, range maps are necessarily a scale‐dependent abstraction ( Hurlbert and White 2005 , Hurlbert and Jetz 2007 ). In general, these maps are typically drawn a smoothed outline around records of occurrence interpolated and can include commission errors or false presences ( Hurlbert and White 2005 ). It should be an initial approach to seed species distributions models and to identify general priorities, which can be refined as more accurate information is obtained for regions with limited information about presence and absence of species ( Lemes et al. 2011 , Rondinini et al. 2011 ). Nevertheless, these maps have been used for modeling species’ ecological niches at broad geographic scales ( Diniz‐Filho et al. 2009 , Lawler et al. 2009 ). We did not use point location data for modeling species’ ecological niches. Although commonly used, point location data are sparse and often biased in their sampling toward areas that are easily accessible, thus increasing omission errors. For these reasons, here we used digital range maps to generate a presence–absence matrix of amphibian occurrence in the Atlantic Forest (see also Lemes and Loyola 2013 ). This matrix, along with climate variables, was then used as our input data for building species’ ecological niche models. Ecological niche models (ENMs) We obtained current climatic data from the WorldClim database (< www.worldclim.org/current >) and future climatic scenarios from CIAT ( ) through WorldClim website. The IPCC's Fourth Assessment Report (AR4) developed these future scenarios. We modeled the ecological niche of each species as a function of four climatic variables: annual mean temperature, temperature seasonality (standard deviation × 100), annual precipitation, and precipitation seasonality (coefficient of variation). These variables are often used to explain geographical/macroecological patterns in amphibian species richness ( Wiens et al. 2006 , Buckley and Jetz 2007 ). Variables for future climate came from three coupled Atmosphere‐Ocean General Circulation Models (AOGCMs), for the A2 scenario for the year 2080: CCCMA‐CGCM2, CSIRO‐MK2 and HCCPR‐HadCM3. The selected AOGCMs are widely used in the literature, having also different equilibrium climate sensitivity values ranging from 3.1°C to 4.4°C. Equilibrium climate sensibility is the annual mean surface air temperature change experienced by the climate system after it has attained a new equilibrium in response to a doubling of CO 2 concentration, and are within the range of all AOGCMs available from IPCC ( IPCC 2007 ). There are several methods for modeling species’ ecological niche and projecting their distributions ( Franklin 2009 ) and there is no consensus on the best method to predict the presence and absence of species according to environmental variables. To cope with these problems, a combination of different projections built upon different conditions and methods – the ensemble forecasting approach ( Araújo and New 2007 ) – proved to be more interpretative than single model analysis. This approach minimizes methodological uncertainties given that the final solution is a unique consensus weighted by the overall statistical fit of combined models, allowing also to quantify and map uncertainties associated with ENMs ( Diniz‐Filho et al. 2009 ). We chose six ENMs: generalized linear models – GLM ( McCullagh and Nelder 1989 ), generalized additive models – GAM ( Yee and Mitchell 1991 ), multivariate adaptive regression splines – MARS ( Muñoz and Felicísimo 2004 ), genetic algorithm for rule set production – GARP ( Stockwell and Noble 1992 ), maximum entropy – MaxEnt ( Phillips et al. 2006 ), and random forest – RF ( Breiman 2001 ). We used both presence‐only and presence–absence methods for modeling species’ ecological niches. Our models can be grouped in two types of methods ( Pearson 2007 ). Methods that use ‘background’ climatic data for the whole study area (e.g. MaxEnt). These methods evaluate how the climatic conditions where species are known to occur relate to the climate across the rest of the study area (the ‘background’); and methods that generate (sample) ‘pseudo‐absences’ from the study area (e.g. GARP). These methods assess differences between the occurrence sites and a set of sites chosen from the study area, which are used instead of real absence data. Here, the set of ‘pseudo‐absences’ were selected randomly ( Stockwell and Noble 1992 ). We evaluated ENMs predictive ability by partitioning our data into training and test sets (at the ratio of 75:25, respectively), with ten replicates of random partitions. To examine the species richness inside PAs and other sites, we established a threshold of pseudo‐absences for each model to allow building the receiving operating curve (ROC) and transform quantitative predictions of models into a binary vector of 0/1, indicating forecasted presences or absences in each grid cell outside species’ extent of occurrences ( Allouche et al. 2006 ). We established the cut‐off point by using the delimitation of the bioclimatic envelope of 95%, allowing us to generate the ROC. Finally, we adopted the True Skill Statistics (TSS) as our measure of model fit. Sensitivity and specificity were calculated based on the probability threshold to which their sum is maximized, not being affected by prevalence. TSS values range from −1 to + 1, where + 1 indicates a perfect fit, minimizing overprediction and omission error rates; values close to −1 indicate a performance worse than those randomly expected ( Allouche et al. 2006 ). We did the ensembles of forecasts to produce more robust predictions and reduce the uncertainties owing to the modeling. We projected distributions to future climate and obtained 180 projections per species within each set of methods (6 modeling methods × 3 AOGCMs × 10 randomly partitioned data) and 60 projections per species for the present (6 modeling methods × 10 randomly partitioned data) – this allowed us to generate a frequency of projections in the ensemble. We then generated the frequency of projections weighted by the TSS statistics for each species and timeframe ( Fig. 1 ). We considered the presence of a species only in cells with 50% or more of frequency of projections, but we held a continuous value when this occurred. This procedure produced two consensus maps, one for the present and one for the future, which were used as input data to analyze amphibian occurrence inside PAs as well as the phylogenetic diversity and composition. 1 Maps of modeled amphibian species richness based on geographic range overlap of 431 species predicted to occur in the Atlantic Forest Hotspot for (A) present and (B) future climatic conditions. Protected areas are shown in black. Amphibian occurrence in PAs We obtained the location of PAs (IUCN categories I–IV) in the Atlantic Forest from the World Database on Protected Areas ( ). We overlaid PA polygons onto our grid considering a grid cell as ‘protected’ even if only a small part of it holds a PA; we assumed therefore that all species occurring in that cell could potentially benefit from the occurrence of a PA in that cell. The incidence of amphibian species in each PA polygon was obtained from consensus maps generated by ENMs based on present or future climatic conditions. Thus, each PA polygon was described by the presence of amphibian species in either present or future climatic conditions. We computed the number of species and Simpson diversity index for each PA polygon under present and future climatic conditions, which was used as a measure of species richness and diversity, respectively. To evaluate the effect of climate change on the number of amphibian species and Simpson diversity, we used paired t‐tests, comparing diversity patterns in the future against present. By doing so, we guaranteed that each PA was compared in terms of present and future amphibian diversity patterns only with itself. Phylogenetic diversity and composition We described each PA polygon by the number of amphibian species per genus for phylogenetic analyses. The phylogenetic tree of amphibian genera from the Atlantic Forest was a pruned version from Pyron and Wiens (2011) . This tree is uses 12 loci from 2871 species. These loci are broadly sampled and have successfully been used in amphibian systematics. A phylogenetic patristic distance matrix ( D F ) for all genera included in Pyron and Wiens’ phylogenetic hypothesis was computed using the software Mesquite 2.73 ( Maddison and Maddison 2010 ) after arbitrary ultrametrization. Posteriorly, we pruned from D F all those genera not occurring in PA polygons, yielding 57 amphibian genera. We also removed seven genera occurring in PA polygons but not included in Pyron and Wiens's tree. We computed phylogenetic diversity using Rao quadratic entropy ( Hardy and Senterre 2007 , Pavoine and Bonsall 2011 ) for each PA polygon under either present or future climatic conditions. To evaluate the effect of climate change on phylogenetic diversity, we used paired t‐tests, comparing diversity patterns in the future against present. We also evaluated phylogenetic composition of amphibian genera using the phylogenetic fuzzy‐weighting method ( Pillar and Duarte 2010 ) and implemented in the R packages SYNCSA ( Debastiani and Pillar 2012 ) and ape ( Paradis et al. 2004 ). This metric uses phylogenetic similarities between taxa to scale‐up the phylogenetic relationships from taxa to the site level. Pairwise phylogenetic distances in D P were transformed into a phylogenetic similarity matrix ( S P = 1 − D P ). Then, phylogenetic similarities in S P were used to weight amphibian genera composition in each PA polygon under present or future climatic conditions, using a fuzzy set algorithm ( Pillar and Duarte 2010 ). This procedure generated a matrix P containing the phylogeny‐weighted genera composition for each PA polygon, under present or future climatic conditions. That is, the presence of genera i in a given PA polygon was shared with each genera j occurring in the whole set of PA polygons (considering both present and future climatic conditions), taking into account the phylogenetic similarity between i and j . Accordingly, those genera j more closely‐related to i (e.g. from the same family) received a proportionally higher fraction of the presence of i in that PA polygon than more phylogenetically distant genera (e.g. from a different order), which will receive a proportionally lower fraction, and so on. Therefore, matrix P expresses the clade composition in PA polygon under present or future climatic conditions. We did a principal coordinates analysis ( Gower 1966 ) on matrix P using square‐rooted Bray–Curtis dissimilarities between PA polygons as resemblance measure to generate principal coordinates of phylogenetic structure (PCPS) for amphibian assemblages. Each PCPS is an eigenvector describing an independent phylogenetic gradient in the dataset ( Duarte 2011 ). The PCPS with the highest eigenvalue describes broader phylogenetic gradients related to deeper tree nodes across protected areas, such as that connecting anurans and caecilians. As the eigenvalues of the other PCPS decrease, finer phylogenetic gradients related to higher nodes (e.g. families, genera) are described ( Duarte et al. 2012 ). PCPS analysis generated 367 eigenvectors with positive eigenvalues. The first four PCPS extracted from matrix P explained > 57% of the total variation in phylogenetic composition in amphibian assemblages ( Table 1 ). 1 Effect of climate change on phylogenetic composition of amphibian genera occurring in protected areas in the Atlantic Forest Biodiversity Hotspot, Brazil. Variable Eigenvalue Contribution to trace (%) t 186 p‐value PCPS 1 2.28 33.64 −31.11 < 0.001 PCPS 2 0.75 11.02 1.82 0.071 PCPS 3 0.54 8.02 −1.30 0.194 PCPS 4 0.31 4.64 1.26 0.209 Trace 6.77 100.00 Each PCPS was analyzed separately using paired t‐tests to evaluate the effect of climate change on phylogenetic composition. PCoA was run in MULTIV 2.63b statistical software (by V. Pillar, available at , user's guide included). Paired t‐tests were performed using SigmaPlot 11 (Systat Software). Results For most species, TSS values were relatively high (TSS ± SD = 0.63 ± 1.33) indicating good model fit. Combined model projections indicated areas of high species richness in central and eastern parts of the Atlantic Forest in the present ( Fig. 1A ). This region should decrease in species richness by 2080, whereas northern and southwestern portions should face many species extinction ( Fig. 1B ). All modeling methods forecasted a general reduction in species’ ranges, leading to a decrease in the number of sites with high species richness. Most species had a significant range contraction (up to 85%) and 12% of species were expected to be regionally extinct (Supplementary material Appendix 1, Table A1). Not surprisingly, climate change had a negative impact on both species richness ( Fig. 2A ) and Simpson diversity ( Fig. 2B ). Nonetheless, phylogenetic diversity increased under future climatic conditions, albeit such increase was not clear ( Fig. 2C ). 2 (A) Number of species, (B) species diversity (Simpson index), and (C) phylogenetic diversity (Rao quadratic entropy) of amphibians modeled according to present and future climatic conditions in Atlantic Forest protected areas. Only PCPS 1 (33% of total variation) showed significant association with climate change, while PCPS 2 (11%) showed marginal significance ( Table 1 ). The occurrence of amphibian genera in PAs under future climatic conditions was negatively related to PCPS 1, while amphibian distribution under present climate was positively related to PCPS 1. Amphibian composition along PCPS 2 in PAs under present climate was negatively associated with this ordination axis, whereas amphibian distribution in the future was positively related to PCPS 2 ( Table 1 ). The ordination scatterplot shows that the occurrence of Gymnophiona, Hemiphractidae, and Pipidae in PAs was positively related to climate change, whereas the occurrence of late‐divergent clades, such as Ceratophryidae, Craugas toridae, Cycloramphidae, Centrolenidae, Eleu therodactylidae, Hylodidae, Hylidae, Microhylidae, and Strabomantidae was negatively associated with climate change in PAs ( Fig. 3 ). Not surprisingly, the distribution of Lithobates , an invasive species in the Atlantic Forest, tended to be positively related to PCPS 1, indicating that climate change will likely favor that species. 3 Scatter diagram of the first two principal coordinates of phylogenetic structure (PCPS) of amphibian genera occurring in Atlantic Forest protected areas. Genera occurrence was modeled according to present (empty squares) and future (grey squares) climatic conditions. Black dots represent species grouped in monophyletic clades in the biplot. Discussion We forecasted shifts and contractions in species’ range for most Atlantic Forest amphibians. On the one hand, most PAs would become climatically unsuitable to sustain their current number of species under climate change, leading to a decrease in species richness and diversity. On the other hand, phylogenetic diversity tended to increase given the higher representativeness of basal clades (Gymnophiona and Pipidae) under future climatic conditions. That is to say that species‐specific and clade‐specific responses to climate change could be largely different, and suggests that amphibian responses to climate change shows a low degree of phylogenetic signal. This is an important message in this paper. This result contradicts the phylogenetic patterns found by Thuiller et al. (2011) . In that study, the authors observed that under a climate change scenario in Europe, mammal, plant, and bird assemblages tended to loose phylogenetic diversity, suggesting that sensitivity to climate change is to some extent phylogeneticaly conserved. Further, our results also provided clues for analyzing the pattern of phylogenetic diversity in a more refined way than provided by other similar methods ( Peres‐Neto et al. 2012 ), given that we were able to visualize how each clade responds to climate change. Families negatively associated with climate change generally have more specialized reproductive modes ( Haddad and Prado 2005 ), such as direct development, with egg laying in humid forest floors (Craugastoridae, Eleutherodactylidae, Strabomantidae, and some Micro hylidae genera), or breed only in primary forest streams (Centrolenidae, Cycloramphidae, Hylodidae). Species of those lineages have narrow habitat requirements for reproduction and foraging, which could make them more prone to extinction or range contractions, likely due to decreased forest cover as a consequence of high temperatures. On the other hand, Pipidae and Gymnophiona tended to increase their representativeness in PAs along the Atlantic Forest. The current distribution of the aquatic genus Pipa in the Atlantic Forest is restricted to warm regions in the northeastern ( Duellman 1999 , Frost 2012 ). With climate change and range shifts, it could expand its distribution southwards, increasing overall phylogenetic diversity of PAs. Other families seem not to be largely affected by climate change, such as Bufonidae and Leptodactylidae. Species of these families are generalists and occur in both open and forested habitats (e.g. Rhinella ) or have reproductive modes independent of water or humid places (e.g. Leptodactylus ; Haddad and Prado 2005 ). Future climate change has also potential implications for the spread of diseases, such as the deadly fungus Batrachochytrium dendrobatidis (Bd) ( Hof et al. 2011 ). It is known that the optimum temperature for Bd growth and development ranges from 17°C to 25°C, with temperatures higher than 30°C being lethal ( Blaustein et al. 2012 ). Therefore, some lineages may benefit from predicted increasing temperatures, since the immune system of amphibians may become more efficient. However, factors affecting susceptibility of species and lineages to infection are still controversial. For example, several factors with a strong phylogenetic pattern seem to play a key role in the resistance to Bd infection ( Blaustein et al. 2012 ), such as peptides and bacteria from amphibian's skin ( Maciel et al. 2003 ), body size, and reproductive mode ( Bancroft et al. 2011 ). Conversely, other studies ( Crawford et al. 2010 ) suggest that a phylogenetic component is not involved in epidemic events of Bd infection in Central America. The relationship between current and future climates and diseases‐driven declines is complex ( Blaustein and Kiesecker 2002 ) and still demands research efforts. Further, our findings indicate that the most likely impact of climatic changes should occur along the coastal line of the Atlantic Forest. This results concur with Lemes and Loyola (2013) and add up to finding of Hof et al. (2011) . These authors found a synergetic effect of climate change, land‐use change and disease spreading in this region. In their study, however, effects of climate change were more pronounced in the north of Atlantic Forest. Menéndez‐Guerrero and Graham (2013) also found synergetic effects of climate change and chytridiomicosis exert great impact on amphibians in Ecuador. These results indicate that climate change might foster infection by pathogens and that management of local populations and disease control are essential for safeguarding amphibians in this regions, in which Bd has been already found ( Eterovick et al. 2005 , Verdade et al. 2012 ). In any case, however, there should be no major changes in species richness due to short latitudinal temperature gradients of the tropics ( Colwell et al. 2008 ). Consequently range shifts will be likely more pronounced toward higher elevations than higher latitudes ( Bush and Hooghiemstra 2005 ). Indeed, our projections for the future predict many species ranges moving upward in elevation, which concentrates a large number of endemic species in the Atlantic Forest ( Carnaval et al. 2009 ), such as brachycephalids that occur in isolated mountaintops (e.g. Brachycephalus pernix ; Pombal Jr et al. 1998 ), and several hylid genera (e.g. Bokermannohyla, Phasmahyla ). In fact, climatically suitable sites tend to concentrate in the mountain range of Serra do Mar, in the eastern part of the biome – exactly where we forecast a high species richness in the future. The whole region is filled with forest refugia from the Pleistocene ( Porto et al. 2012 ) and that might explain why climate is not changing drastically in these portions of the biome. Such reduction of climatically suitable sites leads to an overall range contraction that defines the association of clades to present or future climatic conditions. Clades showing lower range contraction (up to 25%; e.g. Gymnophiona, Pipidae, Supplementary material Appendix 1, Table A1) relate to future climatic conditions. Those with a high range contraction (> 40%; e.g. Aromobatidae, Dendrobatidae, Supplementary material Appendix 1, Table A1) are underrepresented in the future, being related to current climatic conditions. Thus, the increase in phylogenetic diversity is not necessarily good news. It means that, once many clades are loosing climate space in the future, and at the same time, basal clades are being favored, phylogenetic diversity increases in PAs, while species richness decreases. All in all, we are loosing species representation in PAs, while evolutionary information is being maximized by the maintenance of phylogenetically distant clades. Recent studies found similar results on the likely range contraction of the American bullfrog Lithobates catesbeianus ( Giovanelli et al. 2008 , Nori et al. 2011 , Loyola et al. 2012 ), one of the hundred worst invasive species in the world ( Lever 2003 ). As highlighted by Loyola et al. (2012) , a retraction in suitable climatic conditions for L. catesbeianus in western Brazil could drive the alien species into Atlantic Forest PAs. Other studies agree that the bullfrog is expected to increase its representation along Atlantic Forest PAs ( Giovanelli et al. 2008 , Nori et al. 2011 ). Such spread of the species in PAs explains why Lithobates is related to future climatic conditions of PAs in our study. The occurrence of the bullfrog in Atlantic Forest PAs is especially concerning, given that it has deleterious effects on populations of native amphibians and other organisms through competition and predation ( Giovanelli et al. 2008 , Nori et al. 2011 ). As any study based on species distribution models, our analyses have their own caveats. Predictions of range shifts for amphibian species in the Atlantic Forest are clearly dependent of two assumptions: unlimited dispersal ability and absence of biological interactions ( Soberón 2007 ). Model predictions are fraught with uncertainties to obtain range changes ( Diniz‐Filho et al. 2009 ), beyond the stemming ecological and evolutionary lousily represented in the models ( Guisan and Thuiller 2005 , Brook et al. 2009 ). The ensemble forecasting is an alternative and conservative approach for reducing uncertainties among models, which combines the central tendency of multiples bioclimatic niche models ( Araújo and New 2007 ). Consensual models, when appropriately analyzed, may recover species range shifts in climate change context at a broad scale ( Marmion et al. 2009 , Dobrowski et al. 2011 ). Nevertheless, given the uncertain nature of model predictions, our results should be taken with prudence. Further, our models did not include dispersal and adaptation components of species’ distribution. Norberg et al. (2012) , for example, developed an eco‐evolutionary model of multi‐species responses to climate change. They showed that dispersal and adaptation mediate extinction risks in different ways, leading to biodiversity change through time and across changing climates. Integrating these aspects into ecological niche modeling is still in its infancy and new studies in this field are need to allow for applying it for multiple species and broad geographic scales. Incorporating the phylogenetic structure of amphibian assemblages into our analyses show that we should expect a clade‐specific effect of climate change on the distribution of amphibians inhabiting Atlantic Forest PAs. Basal clades should be less affected, while clades retaining specialized reproductive modes should be highly impacted. Identifying major changes in the phylogenetic pool represent a first step towards a better understanding of how assembly processes related to climate change will affect ecological communities. We hope our results contribute to a more deep analysis of the impacts of climate change not only on species, but also on the evolutionary relationships at the local and regional assemblages. Acknowledgements RDL and LDSD received a productivity scholarship awarded by the CNPq. PL is supported by a CNPq PhD scholarship. FTB and DBP receives a CAPES PhD scholarship. The research was funded by the FTC (Brazil/Portugal) program of CAPES, by the Brazilian Network for the study of Climate Change (MCT/Rede CLIMA), and by Conservation International, Brazil.

Journal

EcographyWiley

Published: Jan 1, 2014

There are no references for this article.