Intraspecific Morphological and Genetic Variation of Common Species Predicts Ranges of Threatened Ones Rspb.Royalsocietypublishing.Org Trevon L

Intraspecific morphological and genetic variation of common species predicts ranges of threatened ones rspb.royalsocietypublishing.org Trevon L. Fuller1, Henri A. Thomassen1,†, Manuel Peralvo2, Wolfgang Buermann1,3, Borja Mila´1,4, Charles M. Kieswetter5, Pablo Jarrıń-V6, Susan E. Cameron Devitt7, Eliza Mason8,9, Rena M. Schweizer8, Jasmin Schlunegger8, Janice Chan8, Ophelia Wang10,11, Christopher Research J. Schneider5, John P. Pollinger1,8, Sassan Saatchi1,12, Catherine H. Graham13, 8 1,8 Cite this article: Fuller TL, Thomassen HA, Robert K. Wayne and Thomas B. Smith Peralvo M, Buermann W, Mila´ B, Kieswetter 1Center for Tropical Research, Institute of the Environment and Sustainability, University of California, CM, Jarrıń-V P, Devitt SEC, Mason E, Schweizer Los Angeles, La Kretz Hall, Suite 300, 619 Charles E. Young Dr. East, Los Angeles, CA 90095, USA RM, Schlunegger J, Chan J, Wang O, Schneider 2Iniciativa de Estudios Ambientales Andinos, Consorcio para el Desarrollo Sostenible de la Ecorregioń Andina, CJ, Pollinger JP, Saatchi S, Graham CH, Wayne Diego de Brieda E17-169 y Clemente Celi (Sector Bellavista), PO Box 17-21-1977, Quito, Ecuador 3School of Earth and Environment, University of Leeds, Leeds LS2 9JT, West Yorkshire, UK RK, Smith TB. 2013 Intraspecific morphological 4Museo Nacional de Ciencias Naturales, Consejo Superior de Investigaciones Cientı´ficas, Jose´ Gutie´rrez Abascal 2, and genetic variation of common species Madrid 28006, Spain 5 predicts ranges of threatened ones. Proc R Soc Department of Biology, Boston University, Boston, MA 02215, USA 6Yasuni Research Station, Escuela de Ciencias Biolo´gicas, Pontificia Universidad Cato´lica del Ecuador, Quito, Ecuador B 280: 20130423. 7Department of Wildlife Ecology and Conservation, University of Florida, PO Box 110430, Gainesville, http://dx.doi.org/10.1098/rspb.2013.0423 FL 32611-0430, USA 8Department of Ecology and Evolutionary Biology, University of California, Los Angeles, 621 Charles E. Young Dr. East, Los Angeles, CA 90095, USA 9Department of Microbiology and Immunology, School of Medicine, University of North Carolina, Received: 18 February 2013 804 Mary Ellen Jones Building, Chapel Hill, NC 27599, USA 10Department of Geography and the Environment, Universityof Texas at Austin, Mailcode A3100, Austin, TX 78712, USA Accepted: 22 March 2013 11Center for Sustainable Environments, Northern Arizona University, PO Box 5767, Flagstaff, AZ 86011, USA 12Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Drive, Pasadena, CA 91109, USA 13Department of Ecology and Evolution, Stony Brook University, 650 Life Sciences Building, New York, NY 11794, USA Subject Areas: ecology, environmental science, evolution Predicting where threatened species occur is useful for making informed conservation decisions. However, because they are usually rare, surveying threatened Keywords: species is often expensive and time intensive. Here, we show how regions where common species exhibit high genetic and morphological divergence biodiversity, conservation planning, indicator among populations can be used to predict the occurrence of species of conserva- species, reserve selection, surrogacy tion concern. Intraspecific variation of common species of birds, bats and frogs from Ecuador were found to be a significantly better predictor for the occurrence of threatened species than suites of environmental variables or the occurrence of Author for correspondence: amphibians and birds. Fully 93 per cent of the threatened species analysed had Trevon L. Fuller their range adequately represented by the geographical distribution of the mor- e-mail: [email protected] phological and genetic variation found in seven common species. Both higher numbers of threatened species and greater genetic and morphological variation of common species occurred along elevation gradients. Higher levels of intraspecific divergence may be the result of disruptive selection and/or introgression † Present address: Department of Comparative along gradients. We suggest that collecting data on genetic and morphological Zoology, Institute of Evolution and Ecology, variationincommonspeciescanbeacosteffective tool for conservation plan- University of Tu¨bingen, Auf-der-Morgenstelle ning, and that future biodiversity inventories include surveying genetic and 28, 72076 Tu¨bingen, Germany. morphological data of common species whenever feasible.

Electronic supplementary material is available at http://dx.doi.org/10.1098/rspb.2013.0423 or 1. Introduction via http://rspb.royalsocietypublishing.org. To conserve biodiversity, 168 signatory nations at the 2010 conference of the Convention on Biological Diversity held in Nagoya, Japan agreed to establish

& 2013 The Author(s) Published by the Royal Society. All rights reserved. Table 1. Measures of intraspeciﬁc morphological and genetic variation were developed for seven common species in western Ecuador. 2 sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org genetic markers (loci, individuals, species sites) morphological traits (n, sites)

birds wedge-billed woodcreeper amplified fragment length wing length, tarsus length, tail length, bill length, (Glyphorynchus spirurus) polymorphisms (AFLPs; 136, 178, 15) bill width and bill depth (195, 15) masked flowerpiercer microsatellites (10, 102, 12) wing length, tarsus length, tail length, bill length, (Diglossa cyanea) bill width and bill depth (90, 10) streak-necked flycatcher microsatellites (10, 106, 9) wing length, tarsus length, tail length, bill length, (Mionectes striaticollis) bill width and bill depth (122, 8) bats silky short-tailed bat none centroid size of the skull, angle of curvature of the (Carollia brevicauda) zygomatic arch and forearm length (167, 43) chestnut short-tailed bat none centroid size of the skull, angle of curvature of the (Carollia castanea) zygomatic arch and forearm length (86, 25) seba’s short-tailed bat AFLPs (311, 83, 9) centroid size of the skull, angle of curvature of the (Carollia perspicillata) zygomatic arch and forearm length (160, 44) frog zurucuchu robber frog two anonymous loci and an 840 base- snout-vent length, gape width, lengths of the (Hylodes buergeri) pair region of recombination metacarpal phalanges, length of the radio-ulna, activation gene 1 (RAG1; 3,76–127 lengths of the metatarsal phalanges, length of per locus, 9) the tarsus, tibio–fibula length, femur length and lower jaw length (224, 16)

protected areas comprising 17 per cent of the terrestrial area of number of threatened vertebrates is among the highest in the the planet by 2020 [1]. Because protecting every species is world [9]. infeasible, conservation planners typically focus on threatened species. However, identifying the most critical areas for protecting threatened species requires accurate distributional 2. Material and methods data for these species, most of which are rare and difficult to We sampled genetic and morphological variation in seven survey. This challenge has prompted conservation planners common species in western Ecuador, developed genetic markers to search for surrogates for species of conservation concern and constructed spatial models of genetic and morphological [2]. To be effective, surrogate distributions must be easily divergence among populations of each species. We then designed measured and have a high probability of encompassing areas reserves based on this genetic and morphological variation and where threatened species occur [3]. measured the percentage of threatened species in the reserves. Genetic and morphological variation of common species We also compared reserves designed based on genetic and mor- can be easily measured and are increasingly recognized for phological variation in common species to reserves designed their potential utility in designing protected areas [4]. The abil- based on only environmental variables or occurrences of birds ity to develop and use molecular markers to assess genetic and amphibians (see the electronic supplementary material, variation in non-model organisms is increasingly possible figure S1 summarizes the analysis). The analysis was restricted to western Ecuador because this region of dry forest is particula- owing to advancements in methods and decreases in costs rly threatened by anthropogenic impacts (see the electronic [5]. In addition, measures of intraspecific morphological vari- supplementary material, figure S2 shows our study region). ation for many species can be collected easily in the field or obtained from museum specimens. If genetic and morphological traits of common species are effective for identifying and (a) Measurement of genetic and morphological protecting areas where species of conservation concern occur, traits in common species and generalized these attributes would make them excellent tools to use as indi- cators for conservation. For example, in the Atlantic Forest of dissimilarity modelling Brazil, the genetic divergence of common frogs is an effective Three birds, three bats and one frog species were analysed because predictor of endemism in other taxa [6]. Here, we investigate they were abundant and easily sampled and represent a range of the ability to use genetic and morphological divergence different vagilities and life histories ([4]; table 1). The seven species occupy distinct niches representing a diverse range of altitudes, among populations of seven common species to predict 29 vegetation communities and feeding ecologies, including insectiv- threatened species from western Ecuador. The study region is ory, frugivory and nectarivory. Genetic data were available for a biodiversity hotspot, represents a transition zone between the three birds, one bat and the frog (table 1). We developed anon- humid and dry tropical forests [7] and contains high levels of ymous loci (amplified fragment length polymorphisms (AFLPs), endemic plants and animals, including 650 bird and 6300 vas- microsatellites and nuclear sequence) and targeted loci (recom- cular plant species, 20 per cent of which are endemic [8]. Owing bination activating gene 1 (RAG1)) following published methods to high rates of deforestation and habitat degradation, the [4,10] to assess genetic variation among populations of common species. We used AFLP loci for the wedge-billed woodcreeper of plants to Red-listed species because the exclusive use of the 3 (Glyphorynchus spirurus) [10,11] and seba’s short-tailed bat (Carollia Red List may not adequately represent plants [17], especially in perspicillata; [4]), whereas for the masked flowerpiercer (Diglossa a hotspot of floral endemism such as Ecuador, and may not 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org cyanea) and streak-necked flycatcher (Mionectes striaticollis), we capture the threat of land clearing in dry forest vegetation com- developed microsatellites [4]. Finally, for the zurucuchu robber munities west of the Andes. Instead, we selected plants that frog (Hylodes buergeri), we sequenced two anonymous loci and an were deemed to be of conservation concern based on expert 840 base-pair region of the RAG1 [4]. We then calculated the genetic knowledge and were also representative of vegetation commu- distance between sampled populations. We used Fst as the distance nities that have a high risk of being deforested. Hereafter, the measure for thewedge-billed woodcreeper, streak-necked flycatcher term ‘threatened species’ refers to the 18 plants, five birds and and the masked flowerpiercer because this measure requires genoty- six mammals selected using the aforementioned criteria. pic data, and we had microsatellites and AFLPs for these species. Next, we calculated the number of threatened species occur-

We used Fst for the zurucuchu robber frog because this measure ring in reserves designed based on genetic and morphological uses sequence data, which we had for the frog. variation of common species. We aggregated the occurrence To quantify morphological variation among populations, fol- data to the 100 km2 resolution (see the electronic supplementary lowing published methods [4], we measured morphological material, table S1) because the correlation between surrogates traits on each sampled individual (table 1): for birds, wing, tail, and threatened species is typically highest at this resolution tarsus and bill length, bill width and bill depth; for bats, skull [18]. The resulting dataset consisted of 639 sites that abut each size, zygomatic arch and forearm length; and for the frog, other and occupy a continuous swath of land in western Ecuador snout-vent length, gape width and the length of the metacarpal (see the electronic supplementary material, figure S2). The same phalange, radio-ulna, metatarsal phalange, tarsus, tibio–fibula spatial scale was used for the analyses based on environmental and lower jaw. The frog morphological traits were corrected for variables, threatened species and genetic and morphological size as described elsewhere, but regression analysis indicated traits. In addition to analysing raw data on the known occur- that body size corrections were unnecessary for the birds and rences of threatened species, we also examined models of the bats [4]. The morphological distance between pairs of sampled geographical distribution of threatened species, which predict populations was calculated as the Euclidean distance between species’ niches using occurrences and environmental variables. the mean phenotype in each population. The total number of The models used here were constructed by Peralvo et al. [13] genetic and morphological traits for the seven common species with the GARP software package using environmental data com- was 41 (36 morphological and five genetic traits; table 1). prising temperature and precipitation variables, elevation and We used generalized dissimilarity modelling (GDM), a vegetation type. We calculated the number of threatened species matrix regression technique [12], to identify environmental that had 10 per cent of their predicted distributions included in variables that were significantly correlated with genetic and reserves designed based on genetic and morphological variation morphological distance. We used the implementation of GDM of common species. for the ArcView, and SPLUS software packages to predict value of each of the 36 genetic and morphological traits across western Ecuador. For each trait, we assessed the importance of 14 inde- (c) Modelling the geographical range of common birds pendent variables that represent climate, vegetation phenology derived from satellite data and elevation [4,10]. The contributions and amphibians of independent variables to explaining each trait were tested Since the occurrence of common species are effective surrogates in by permutation (models for 36 traits performed better than some geographical regions [19], we also calculated the number of random [4,10]). threatened species in reserves designed based on occurrences of common birds and amphibians. This aspect of the analysis did not incorporate genetic and morphological variation within common species. We computed occurrences of common birds (b) Species of conservation concern and amphibians in western Ecuador from maps available from Our dataset on species of conservation concern comprised a total the NatureServe INFONATURA database (www.natureserve. of 29 plants and vertebrates. Peralvo et al. [13] detail the species org/infonatura; [20–22]). The dataset comprised 1397 amphibian selection criteria, which we will summarize here. For vertebrates, and bird species (Columbiformes and suboscine passerines) we started with 315 species classified as conservation targets by mapped at a 1 km resolution, which we aggregated to 100 km2. TNC-Ecuador’s technical staff because the species are under For each 100 km2 site, the dataset indicated whether each of the threat and restricted to the Equatorial Pacific ecoregion of wes- 1397 species was present or absent in the site. tern Ecuador and northwestern Peru. The list was filtered to retain species that were in the 2012 IUCN Red List and had at least 10 records in our vertebrate occurrence database [13]. The (d) Modelling environmental variables refined list comprised five birds and six mammals (table 2). We Certain climatic, edaphic and topographic features can be effective required 10 or more occurrences because we constructed geo- surrogates for predicting the occurrence of some threatened graphical distribution models for each species (see below), and species [23]. We therefore compared the predictive performance the modelling technique that we used needed 10 observations of genetic and morphological divergence with that of environmen- to make accurate predictions [14]. For plants, we began with tal variables. Environmental variables (aspect, climate, elevation TNC-Ecuador’s list of vascular plants of conservation concern, and slope) available at the global scale were obtained from which comprises 1050 species. Botanists specializing in Ecuador online global databases and clipped to our study region (see the (B. Leo´n and D. Padilla) identified one diagnostic species that electronic supplementary material, table S3). With the exception was representative of each vegetation community in the Equator- of soil type, the environmental variables were continuous. We ial Pacific (communities were defined as per [15]). To keep the discretized each continuous variable into several categorical vari- analysis tractable, we filtered the list to retain just these diagnos- ables. In general, the effectiveness of an environmental variable tic species. Last, we excluded diagnostic species with less than as a surrogate for threatened species is expected to increase with 10 records in the Missouri Botanical Garden’s VAST database the number of categories. This is because a large number of cat- [13]. This resulted in a list of 18 plants (table 2), all of which egories tend to capture fine-scale environmental variation in the are restricted to the Equatorial Pacific and five of which are in sites occupied by the species. However, the computational diffi- Ecuador’s national plant Red List [16]. We did not limit our list culty of the area prioritization problem also increases with the Table 2. Ecuadorian species of conservation concern included in the analysis (D, decreasing; S, stable; U, unknown; MOBOT, Missouri Botanical Garden).

IUCN Red IUCN population Ecuador’s plant MOBOT plants family scientiﬁc name common name occurrences List trend Red List of Ecuador database

birds Cracidae Ortalis erythroptera rufous-headed 16 X D chachalaca Fringillidae Carduelis siemiradzkii saffron siskin 15 X D Picidae Campephilus guayaquil woodpecker 24 X D gayaquilensis Psittacidae Aratinga erythrogenys red-masked parakeet 27 X D Pionus chalcopterus bronze-winged parrot 15 X U mammals Canidae Pseudalopex sechurae sechuran zorro 21 X U Felidae Leopardus pardalis ocelot 45 X D Puma concolor puma 28 X D Muridae Oryzomys xantheolus yellowish rice rat 35 X S Phyllostomidae Artibeus fraterculus fraternal fruit-eating bat 40 X D Sciurus stramineus guayaquil squirrel 51 X U vascular Fabaceae Albizia multiﬂora none 28 U X plants Caesalpinia glabrata none 25 X Centrolobium ochroxylum amarillo 16 U X X Geoffroea spinosa almendon 24 U X Leucaena trichodes capra 45 U X Machaerium millei chiche 27 U X Pithecellobium excelsum chaquiro 28 U X Swartzia haughtii none 38 X D X X Anacardiaceae Mauria heterophylla fenogreco 23 U X Araceae Anthurium none 42 U X X dolichostachyum Bombacaceae Eriotheca ruizii pasallo 19 U X Boraginaceae Cordia lutea yellow geiger 26 U X Capparaceae Capparis heterophylla caper shrub 17 X D X X

(Continued.) sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org 4 number of categories. Sarkar et al. [18] analysed the effect of 5 the number of categories on the effectiveness of environmental surrogates for threatened species on three continents, and identified 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org the optimal number of classes into which climatic and topographic variables should be partitioned to represent threatened species effectively while remaining computationally tractable [18,24]. Our analysis uses their partitioning scheme, which divides each

MOBOT plants of Ecuador database climatic variable into 4–10 equal interval classes (see the electronic supplementary material, table S3). Next, we designed reserves to include at least 10 per cent of the sites in each class of each environmental variable. We then compared the models of the predicted distributions of the threatened species with these reserves. We determined the percentage of threatened species that had 10 per cent of their predicted distributions included in reserves designed

Ecuador’s plant Red List based on the environmental variables.

(e) Assessing the effectiveness of surrogates for threatened species when reserves are prioritized without areal constraints We used a complementarity-based algorithm in the RESNET v. 1.2 software package [25] to select sites containing each surrogate IUCN population trend (genetic and morphological traits of common species, occurrences of common amphibians and birds and environmental variables). Complementarity chooses sites iteratively such that the site selected in each iteration contains the largest number of surrogates not included at the targeted levels in sites chosen in previous iter- ations ([3]; The targeted level is the percentage of the surrogates’

IUCN Red List occurrences that we want to include in reserves.). Since reserve selection algorithms require discrete data [26], we divided the GDM predictions about genetic and morphological divergence into 50 distinct categories per trait (the use of a slightly different number of classes yields similar results). Fifty were selected so that the reserve selection algorithm would remain computationally tractable. We required the selected sites to satisfy a conservation target of 10 per cent for each surrogate [26], meaning that the reserves included at least 10 per cent of the occurrences of each surrogate. We selected 10 per cent because this target has seen widespread use in conservation planning exercises [18,24], but the use of higher targets gave similar results (see the electronic supplementary material, figure S4). However, in this stage of the analysis, there was no constraint on the amount of land in western Ecuador that could be included in the selected sites.

andirobanonelong johnnonebastard cedar 36 32 21 22 26After X designing U reserves, D U Uwe U compared thenumber X of threa- X X X X X tened species included in sites selected based on genetic and morphological variation of common species, environmental variables and occurrences of amphibian and bird species using a one- tailed, unequal variance t-test. If we had used a two-tailed test, the rejection of the null hypothesis would have meant that the number of threatened species is different in areas selected based on genetic and morphological traits, and areas selected based on environmental variables or occurrences of common species without indicating the direction of the difference. Since we hypothesized Carapa guianensis Sorocea sarcocarpa Triplaris cumingiana Solanum confertiseriatum Guazuma ulmifolia that there would be more threatened species at sites selected based on genetic and morphological variation of common species, we used a one-tailed test. The rejection of the null hypothesis in a one-tailed test implies that the number of threatened species in areas prioritized based on genetic and morphological variation is greater than the number of threatened species in sites selected Meliaceae Moraceae Polygonaceae Solanaceae Sterculiaceae family scientiﬁc name common name occurrences based on environmental variables or occurrences of common

) birds and amphibians. The test assumed unequal variance because the number of sites selected based on environmental variables, species occurrences and genetic and morphological variation of Continued.

( common species were different. Owing to the unequal sample sizes, the unequal variance test was the most appropriate because it adjusts the test statistic and degrees of freedom based on the

Table 2. number of samples in the two groups [16]. 20 6 sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org

CI) 12 %

8 (mean ± 95

no. threatened species included in the selected sites 0 (i) (ii) (iii) (iv) Figure 1. The mean number of threatened and endangered species (+95% CI) in the sites selected based on: (i) genetic and morphological traits of seven common species, (ii) environmental variables, (iii) occurrences of amphibians and birds, and (iv) randomly selected sites (Â10 000). The average number of threatened and endangered species in sites selected based on genetic and morphological traits of common species is significantly greater than the number in sites selected based on environmental variables, occurrences of amphibians and birds or a null model consisting of sites selected at random.

(f) Assessing the effectiveness of surrogates for the sites selected based on the surrogate. The y-coordinate of each point is the percentage of the 29 threatened species that are threatened species when reserves are prioritized included in sites selected subject to the areal constraint. When a curve representing a surrogate is above the curve for random subject to areal constraints sites, the surrogate does better than random. When a curve for Since land cost is an important determinant of which sites may one surrogate is above the curve for another surrogate, the first be appropriate for inclusion in a conservation plan [27,28], and surrogate is more effective than the second surrogate. analysing cost makes our planning exercise more realistic, we incorporated a budgetary ceiling on the total area of reserves selected based on genetic and morphological traits of common species, occurrences of common birds and amphibians and environmental variables (see the electronic supplementary mate- 3. Results rial analyses other measures of land cost, including proximity to (a) The effectiveness of genetic and morphological infrastructure). For each surrogate, we constructed a curve called a surrogacy curve using the SURROGACY software package [29,30]. traits as surrogates for threatened species when A surrogacy curve is a generalization of a species accumulation curve that lists how the number of threatened species increases reserves are prioritized without areal constraints as more land is selected based on a surrogate [23]. Each curve When we designed reserves without restricting the amount of describes the properties of 17 different reserve networks selected land that could be put under a conservation plan, sites based on the surrogate. We first selected reserves based on a sur- selected based on genetic and morphological divergence of rogate subject to the constraint the reserves could only occupy 1 common species contained a higher proportion of threatened per cent of the land in western Ecuador and calculated the species than sites selected at random or regions selected using percentage of threatened and endangered species in the selected environmental variables or bird and amphibian occurrences sites. Next, we selected sites based on a surrogate subject to the con- as a proxy (figure 1). In particular, sites selected based on straint that the sites could only occupy 2 per cent of western genetic and morphological traits contain 2.1 times as many Ecuador and calculated the percentage of threatened and endan- threatened species as sites selected based on environmental gered species that were included in the selected sites. The procedure was repeated at 1 per cent increments up to 17 per cent variables, and 1.7 times as many as sites based on bird and of the land in western Ecuador. The upper target of 17 per cent amphibian occurrences (see the electronic supplementary was used because the parties to the Convention on Biological material, figure S2). The number of threatened and endangered Diversity agreed to establish protected areas comprising 17 per is significantly greater than in sites selected based on bird cent of the terrestrial area of the planet [1]. This target is arbitrary and amphibian occurrences (one-tailed, unequal variance to the extent that one should protect whatever per cent of a species’ t-test, t ¼ 6.495, d.f. ¼ 116.043, p ¼ 1.076 Â 1029) or environ- range is sufficient to ensure its survival. However, owing to the dif- mental variables (one-tailed, unequal variance t-test, t ¼ 2.102, ficulty of estimating this percentage, the 17 per cent target is used d.f. ¼ 171.355, p ¼ 0.0185). Sites chosen based on genetic and by many countries [31]. morphological variation of common species contained steeper To provide a comparison with the three surrogate sets (genetic elevational gradients than unselected sites in the study region and morphological traits, occurrences of common species or (one-tailed, unequal variance t-test, t ¼ 3.582, d.f. ¼ 14.409, environmental variables), we also randomly selected sites 10 000 ¼ times. The ‘random’ curve in the surrogacy plot shows the mean p 0.00144). These steep altitudinal gradients also contained number of threatened and endangered species included in sites a high number of threatened species. Finally, we found signi- chosen randomly at targets ranging from 1 to 17 per cent. In the ficant spatial overlap between reserves selected based on curves representing the other surrogates in the surrogacy plot, genetic and morphological traits of common species and the x-coordinate of each point is the areal constraint (1–17%) on models of the geographical distributions of threatened species 100 7 (a) geographic distribution models (b) species occurrences sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org

threatened species in protected areas (%) areas priorititized to include genetic and 20 morphological traits of 7 common species areas priorititized to include environmental surrogates areas priorititized to include occurrences of amphibians and birds (1397 species) random area prioritizations 0 2.5 5.0 7.5 10.0 12.5 15.00 2.5 5.0 7.5 10.0 12.5 15.0 area selected to be protected (%) area selected to be protected (%) Figure 2. Effectiveness of genetic and morphological surrogates for threatened species. (a) Habitat models. Uses models of the geographical distributions of threatened species. The solid black line with white circles represents the percentage of threatened species that is found in areas designed based on unique intraspecific morphologic and genetic divergence of common species (areas priorititized to include genetic and morphological traits of seven common species). The solid black line with white triangles depicts areas prioritized based on occurrences of birds and amphibians (areas priorititized to include environmental surrogates). The solid black line with squares represents environmental surrogates (areas priorititized to include occurrences of amphibians and birds (1397 species)). The dashed grey line shows the percentage of threatened species in random areas (random area prioritizations). We carried out 10 000 randomizations for each point, and calculated the mean and 99% CIs (the confidence intervals were too small to be depicted). Since the black line with white circles is to the left of the other lines, areas designed based on intraspecific variation of common species protect the highest percentage of threatened species. (b) Species occurrences. Uses raw data on the occurrences of threatened species rather than geographical distribution models. The ‘random’ lines differ in (a)and(b) because they are based on different areas prioritized at random.

(Cramer–Von Mises test: null hypothesis of significant overlap, the common species we surveyed performed better than p . 0.05; electronic supplementary material, table S2). random, though the benefit of the method is limited as one approaches the 17 per cent target (figure 2a). Second, we compared reserves selected based on genetic and morphological (b) The effectiveness of genetic and morphological traits of common species with known occurrences of threatened traits as surrogates for threatened species when species. We found that reserves selected based on genetic and morphological traits include a significantly higher percentage reserves are prioritized subject to areal constraints of threatened species than sites selected at random. In the corre- When the maximum amount of land that could be put under a sponding surrogacy plot, the curve that represents genetic and conservation plan ranged from 1 to 17 per cent of western morphologicaltraitsisconsistently higher than the curve for Ecuador, intraspecific variation of the seven common species random sites (figure 2b). In addition, when we incorporated remained the most effective surrogate for species of conservation land cost into the prioritizations, sites selected based on intra- concern. First, we compared the reserves selected based on gen- specific genetic and morphological divergence included etic and morphological data of common species with models of significantly more threatened species than random sites or sites the geographical distributions of threatened species. We found prioritized based on environmental parameters or bird and that genetic and morphological traits of the seven common amphibian occurrences (see the electronic supplementary species were the most effective surrogate for threatened species material, figure S3). because the curve that represents genetic and morphological traits is consistently higher than the curves for random sites, environmental variables or bird and amphibian occurrences 4. Discussion (figure 2a). Of the 29 threatened species analysed, 27 have at (a) Processes that generate divergence and threatened least 17 per cent of their ranges represented in regions identified based on intraspecific morphological, and genetic variation species along gradients exhibitedbythesevencommonspecies.Ineverycase,regions Our results show that the genetic and morphological vari- selected to represent genetic and morphological divergence in ation exhibited by common species effectively predicts the occurrence of threatened species in western Ecuador. In We hypothesize that elevational gradients, because they are 8 particular, when we prioritized reserves without areal con- harder to reach than lowlands, contain some of the last sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org straints, we found that elevational gradients supported forest remnants west of the Andes but that deforestation is higher intraspecific divergence of genetic and morphological expanding to steeper areas. When we analysed a satellite-derived traits in common species and higher numbers of threatened map of western Ecuador’s land cover [48], we found that the species than other sites, such as lowlands. Genetic and mor- proportion of land recently converted to cropland and artificial phological traits of common species may predict the ranges areas is greater along steep elevational gradients than in flat threatened species well simply because the two overlap areas (one-tailed t ¼ 5.6109, d.f. ¼ 19.953, p ¼ 1.044 Â 1025). spatially along gradients of elevation. Why should this be This could be due to the fact that flat areas were largely cleared the case? Below we examine the processes acting along decades ago, or owing to the expansion of infrastructure in gradients that may help explain patterns. steep areas such road development that results in deforestation Elevational gradients typically have substantial ecological of previously inaccessible sites. For example, in the Atlantic turnover because some ecological communities are restricted Forest of Brazil, a disproportional amount of road construction to a small altitudinal range [4,10,32]. The high number of in steep areas from the 1960s to the 1980s contributed signifi- communities along an elevational gradient could promote cantly to forest fragmentation [49]. If roads have expanded in diversifying selection as populations adapt to differing abio- steep areas in Ecuador in a similar manner, threatened spe- tic and biotic conditions. This could generate high levels of cies may concentrate along gradients, because gradients are intraspecific genetic and morphological variation within undergoing high rates of road-associated land clearing. common species. For example, since temperature and precipi- Although testing this hypothesis remains an important tation generally decrease with elevation, selection could lead area for future research, findings from elevational gradients to both phenotypes adapted to dry, cold environments at the in other regions may provide insight into whether deforesta- higher ends of gradients and hot, wet environments towards tion could have led to a high number of threatened species the lowlands [33,34]. The outcome would be high levels of vari- along elevational gradients in western Ecuador. For example, ation in morphological traits related to thermal and water deforestation along elevational gradients in the Columbian tolerance across the gradient. Such patterns of selection Andes during the first half of the twentieth century resulted across ecological gradients may ultimately lead to reproductive in a significant decrease in insect diversity [50]. Furthermore, isolation and speciation [35–38]. in the Sierra Nevada mountain range of California during the Elevational gradients may also be hotbeds of genetic diver- past century, animal populations occurring along altitudinal gence within common species because they undergo strong gradients have experienced range contraction owing to cyclical fluctuations in the environment. Quantitative genetic anthropogenic effects such as land clearing and climate models predict that when the environment varies in a cyclical change [51,52]. A similar process may have occurred in manner, having high additive genetic variance will signifi- steep areas in western Ecuador, leading to a high number cantly increase a population’s average fitness [39]. Over of threatened species along altitudinal gradients. decadal time scales, there is significant variation in temperature along elevational gradients in Ecuador that is triggered by the El Nin˜ o–Southern Oscillation, and this variation is (b) Implications for biodiversity monitoring more pronounced along gradients than at lower altitudes Surveys of biodiversity in remote areas often have the goal [40]. In the light of this, populations that diverge genetically of inventorying all species in a region. However, such as a means of adapting to cyclically varying climate along approaches are costly because they require that experts be elevational gradients may have higher fitness, which could dispatched to survey multiple taxonomical groups in the result in high intraspecific genetic variation along gradients. field and funding for subsequent analyses. Our analyses Moreover, introgression, hybridization between species fol- show that genetic and morphological traits of common lowed by backcrossing with parentals, might also result in species may be used as effective surrogates for the occurrence genetic divergence along elevational gradients because hybrids of species of conservation concern. Given that they are effec- carry alleles from both parental lineages [41]. Since introgres- tive surrogates, future rapid assessments of biodiversity sion has been shown to occur along elevational gradients might be improved or supplemented by also sampling when high- and low-altitude species hybridize at intermediate amphibian, avian and chiropteran genetic and morphological altitudes [42–44], elevational gradients in western Ecuador divergence of selected common species. This has the potential may have high genetic divergence among populations of to significantly decrease the time and cost of biodiversity common species because these gradients are hybrid zones. assessments, making it possible to carry out a greater Finally, because elevational gradients have high ecologi- number of assessments each year. Lastly, surveying genetic cal turnover, species along the gradient may exhibit high and morphological variation could lead to direct tests about levels of specialization [45]. These specialist species would the processes resulting in biodiversity and how it may be confined to relatively small ranges, because only a respond to climate change [53,54]. narrow elevational band is ecologically suitable for them, Since our conclusion that intraspecific variation of birds, and consequently they may be more easily threatened since bats and a frog is an effective surrogate for threatened species extinction risk decreases with range size [46]. Anthropogenic is based on analysis of only seven species, this raises the ques- disturbance would further increase the risk of extinction of tion of how generalizable our conclusions are. A search of the rare, specialized species along gradients. An example of literature reveals that intraspecific diversity has also been such disturbance is deforestation. Significant clearing of for- shown to be correlated with diversity and endemism at the ests in western Ecuador began in the 1950s [8]. The main species level in a number of other taxa and geographical drivers of deforestation in this region have been the expan- regions, including birds in the West Indies [55], butterflies in sion of cattle pasture, cropland and timber extraction [47]. Indonesia [56], stream fishes in the midwestern US [57] and beetles in the Aegean archipelago [58]. The association between samples from common species and analysing genetic and mor- 9 species diversity and intraspecific variation in diverse taxa phological data is less expensive than surveys to detect sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org suggests that our results may be useful for other taxa. However, threatened species, then our method will have practical utility future work should test whether the association between for governmental agencies and conservation NGOs. Compar- genetic and morphological traits of common species and threa- ing the cost-efficiency of these two methods remains an tened species is limited to areas where there are steep altitudinal important area for future research. gradients. Given the relative ease associated with assaying genetic variation in non-model organisms, we believe the We thank R. Calsbeek, R. Harrigan, B. Larison, R. E. Ricklefs, three reviewers and Associate Editor Daniel Rabosky for comments that approach presented here represents a potentially important improved the manuscript. This work was supported by NSF grant new tool for the conservation of biodiversity. Whether this no. IRCEB9977072 to T.B.S, R.K.W. and C.J.S.; and NASA grant nos tool can be implemented in real-world conservation planning IDS/03–0169–0347 to T.B.S., R.K.W., and C.J.S.; and NNG05GB37G will depend largely on economic cost: if collecting genetic to C.H.G.

References

1. Convention on Biological Diversity. 2010 Strategic and predict patterns of beta diversity in regional 23. Faith DP, Walker PA. 1996 Environmental diversity: plan for biodiversity, 2011–2020. Montreal, biodiversity assessment. Divers. Distrib. 13, on the best possible use of surrogate data for Canada: Secretariat of the CBD. 252–264. (doi:10.1111/j.1472-4642.2007.00341.x) assessing the relative biodiversity of sets of areas. 2. Caro TM. 2010 Conservation by proxy. Washington, 13. Peralvo M, Sierra R, Young KR, Ulloa-Ulloa C. 2007 Biodivers. Conserv. 5, 399–415. (doi:10.1007/ DC: Island Press. Identification of biodiversity conservation priorities BF00056387) 3. Sarkar S et al. 2006 Biodiversity conservation using predictive modeling: an application for the 24. Sarkar S, Mayfield M, Cameron S, Fuller T, Garson J. planning tools: present status and challenges for Equatorial Pacific region of South America. Biodivers. 2007 Conservation area networks for the Indian the future. Ann. Rev. Environ. Res. 31, 123–159. Conserv. 16, 2649–2675. (doi:10.1007/s10531- region: systematic methods and future prospects. (doi:10.1146/annurev.energy.31.042606.085844) 006-9077-y) Himalayan J. Sci. 4, 27–40. (doi:10.3126/ 4. Thomassen HA et al. 2011 Mapping evolutionary 14. Peterson AT. 2001 Predicting species’ geographic hjs.v4i6.980) process: a multi-taxa approach to conservation distributions based on ecological niche modeling. 25. Sarkar S, Fuller T, Aggarwal A, Moffett A, Kelley C. prioritization. Evol. Appl. 4, 397–413. (doi:10.1111/ Condor 103, 599–605. (doi:10.1650/0010- 2009 The ConsNet software platform for systematic j.1752-4571.2010.00172.x) 5422(2001)103[0599:PSGDBO]2.0.CO;2) conservation planning. In Spatial conservation 5. Hayden S, Bekaert M, Crider TA, Mariani S, Murphy 15. Ceron C, Palacios W, Valencia R, Sierra R. 1999 Las prioritization: quantitative methods and WJ, Teeling EC. 2010 Ecological adaptation formaciones naturales de la costa del Ecuador. In computational tools (eds A Moilanen, KA Wilson, determines functional mammalian olfactory Propuesta preliminar de un sistema de clasificacion HP Possingham), pp. 235–248. New York, NY: subgenomes. Genome Res. 20, 1–9. (doi:10.1101/ de vegetacion para el Ecuador Continental Oxford University Press. gr.099416.109) (ed. R Sierra), pp. 55–78. Quito, Ecuador: INEFAN/ 26. Margules CR, Sarkar S. 2007 Systematic conservation 6. Carnaval AC, Hickerson MJ, Haddad CFB, Rodrigues GEF-BIRF and EcoCiencia. planning. Cambridge, UK: Cambridge University MT, Moritz C. 2009 Stability predicts genetic 16. Valencia R, Pitman N, Le´on-Ya´nez S, Jorgensen PM. Press. diversity in the Brazilian Atlantic Forest hotspot. 2000 Libro Rojo de las Plantas Ende´micas del 27. Ando A, Camm J, Polasky S, Solow A. 1998 Species Science 323, 785–789. (doi:10.1126/science. Ecuador. Quito, Ecuador: Pontificia Universidad distributions, land values, and efficient 1166955) Cato´lica del Ecuador. conservation. Science 279, 2126–2128. (doi:10. 7. Myers N, Mittermeier RA, Mittermeier CG, da 17. Butchart SHM, Stattersfield AJ, Baillie J, Bennun LA, 1126/science.279.5359.2126) Fonseca GAB, Kent J. 2000 Biodiversity hotspots for Stuart SN, Akcakaya HR, Hilton-Taylor C, Mace GM. 2005 28. Bode M, Wilson KA, Brooks TM, Turner WR, conservation priorities. Nature 403, 853–858. Using Red List Indices to measure progress towards the Mittermeier RA, McBride MF, Underwood EC, (doi:10.1038/35002501) 2010 target and beyond. Phil. Trans. R. Soc. B 360, Possingham HP. 2008 Cost-effective global 8. Dodson CH, Gentry AH. 1991 Biological extinction in 255–268. (doi:10.1098/rstb.2004.1583) conservation spending is robust to taxonomic group. western Ecuador. Ann. Mo. Botanical Garden 78, 18. Sarkar S, Justus J, Fuller T, Kelley C, Garson J, Proc. Natl Acad. Sci. USA 105, 6498–6501. (doi:10. 273–295. (doi:10.2307/2399563) Mayfield M. 2005 Effectiveness of environmental 1073/pnas.0710705105) 9. Hoffmann M et al. 2010 The impact of conservation surrogates for the selection of conservation area 29. Garson J, Sarkar S. 2002 Surrogacy manual ver 1.1. on the status of the world’s vertebrates. Science networks. Conserv. Biol. 19, 815–825. (doi:10. Austin, TX: Biodiversity and Biocultural Conservation 330, 1503–1509. (doi:10.1126/science.1194442) 1111/j.1523-1739.2005.00236.x) Laboratory, Section of Integrative Biology, University 10. Thomassen HA et al. 2010 Modeling 19. Ricketts TH. 2001 Aligning conservation goals: are of Texas at Austin. environmentally associated morphological and patterns of richness and endemism concordant 30. Garson J, Aggarwal A, Sarkar S. 2002 Birds as genetic variation in a rainforest bird, and its at regional scales? Anim. Biodivers. Conserv. 24, surrogates for biodiversity: an analysis of a data set application to conservation prioritization. Evol. Appl. 91–99. from southern Quebec. J. Biosci. 27, 347–360. 3, 1–16. (doi:10.1111/j.1752-4571.2009.00093.x) 20. IUCN CI, NatureServe. 2008 An analysis of (doi:10.1007/BF02704965) 11. Mila B, Wayne RK, Fitze P, Smith TB. 2009 amphibians on the 2008 IUCN Red List. Gland, 31. Rodriguez-Rodriguez D, Bomhard B, Butchart SHM, Divergence with gene flow and fine-scale Switzerland: IUCN. Foster MN. 2011 Progress towards international phylogeographical structure in the wedge-billed 21. NatureServe. 2004 InfoNatura: birds, mammals, and targets for protected area coverage in mountains: woodcreeper, Glyphorynchus spirurus, a Neotropical amphibians of Latin America. Arlington, VA: a multi-scale assessment. Biol. Conserv. 144, rainforest bird. Mol. Ecol. 18, 2979–2995. (doi:10. NatureServe. 2978–2983. (doi:10.1016/j.biocon.2011.08.023) 1111/j.1365-294X.2009.04251.x) 22. Ridgely RS, Greenfield PJ. 2001 The birds of 32. Hall J, Burgess ND, Lovett J, Mbilinyi B, Gereau RE. 12. Ferrier S, Manion G, Elith J, Richardson K. 2007 Ecuador. Volume II: a field guide. London, UK: 2009 Conservation implications of deforestation Using generalized dissimilarity modelling to analyse Christopher Helm. across an elevational gradient in the Eastern Arc Mountains, Tanzania. Biol. Conserv. 142, 41. Smith PF, Konings A, Kornfield I. 2003 Hybrid origin 50. Medina CA, Escobar F, Kattan GH. 2002 Diversity and 10 2510–2521. (doi:10.1016/j.biocon.2009.05.028) of a cichlid population in Lake Malawi: implications habitat use of dung beetles in a restored Andean sbryloitpbihn.r rcRScB20 20130423 280: B Soc R Proc rspb.royalsocietypublishing.org 33. Luoto M. 2000 Modelling of rare plant species for genetic variation and species diversity. Mol. Ecol. landscape. Biotropica 34, 181–187. richness by landscape variables in an agriculture 12, 2497–2504. (doi:10.1046/j.1365-294X.2003. 51. Rubidge EM, Patton JL, Lim M, Burton AC, Brashares area in Finland. Plant Ecol. 149, 157–168. 01905.x) JS, Moritz C. 2012 Climate-induced range (doi:10.1023/A:1026531400356) 42. Culumber ZW, Fisher HS, Tobler M, Mateos M, contraction drives genetic erosion in an alpine 34. Lomolino MV. 2001 Elevation gradients of species- Barber PH, Sorenson MD, Rosenthal GG. 2011 mammal. Nat. Clim. Change 2, 285–288. (doi:10. density: historical and prospective views. Glob. Ecol. Replicated hybrid zones of Xiphophorus swordtails 1038/nclimate1415) Biogeogr. 10, 3–12. (doi:10.1046/j.1466-822x. along an elevational gradient. Mol. Ecol. 20, 52. Tingley MW, Koo MS, Moritz C, Rush AC, Beissinger 2001.00229.x) 342–356. (doi:10.1111/j.1365-294X.2010.04949.x) SR. 2012 The push and pull of climate change 35. Smith TB, Wayne RK, Girman DJ, Bruford MW. 1997 43. Floate KD. 2004 Extent and patterns of causes heterogeneous shifts in avian elevational A role for ecotones in generating rainforest hybridization among the three species of Populus ranges. Glob. Change Biol. 18, 3279–3290. (doi:10. biodiversity. Science 276, 1855–1857. (doi:10. that constitute the riparian forest of southern 1111/j.1365-2486.2012.02784.x) 1126/science.276.5320.1855) Alberta, Canada. Can. J. Bot. 82, 253–264. (doi:10. 53. Etterson JR, Shaw RG. 2001 Constraint to adaptive 36. Freedman AH, Thomassen HA, Buermann W, Smith 1139/b03-135) evolution in response to global warming. Science TB. 2010 Genomic signals of diversification along 44. Paige KN, Capman WC. 1993 The effects of host- 294, 151–154. (doi:10.1126/science.1063656) ecological gradients in a tropical lizard. Mol. Ecol. plant genotype, hybridization, and environment on 54. Gienapp P, Teplitsky C, Alho JS, Mills JA, Merila J. 19, 3773–3788. (doi:10.1111/j.1365-294X.2010. gall-aphid attack and survival in cottonwood: the 2008 Climate change and evolution: disentangling 04684.x) importance of genetic studies and the utility of environmental and genetic responses. Mol. Ecol. 17, 37. Smith TB, Calsbeek R, Wayne RK, Holder KH, RFLPs. Evolution 47, 36–45. (doi:10.2307/2410116) 167–178. (doi:10.1111/j.1365-294X.2007.03413.x) Pires D, Bardeleben C. 2005 Testing alternative 45. Fjeldsa J. 1994 Geographic patterns for relict and 55. Vellend M. 2003 Island biogeography of genes mechanisms of evolutionary divergence in an young species of birds in Africa and South America and species. Am. Nat. 162, 358–365. (doi:10.1086/ African rain forest passerine bird. J. Evol. and implications for conservation priorities. Biodivers. 377189) Biol. 18, 257–268. (doi:10.1111/j.1420-9101.2004. Conserv. 3, 207–226. (doi:10.1007/BF00055939) 56. Cleary DFR, Fauvelot C, Genner MJ, Menken SBJ, 00825.x) 46. Manne LL, Brooks TM, Pimm SL. 1999 Relative risk Mooers AO. 2006 Parallel responses of species and 38. Schneider CJ, Smith TB, Larison B, Moritz C. 1999 A of extinction of passerine birds on continents and genetic diversity to El Nino Southern Oscillation- test of alternative models of diversification in islands. Nature 399, 258–261. (doi:10.1038/20436) induced environmental destruction. Ecol. Lett. 9, tropical rainforests: ecological gradients vs. 47. Sierra R, Stallings J. 1998 The dynamics and social 301–307. (doi:10.1111/j.1461-0248.2005.00876.x) rainforest refugia. Proc. Natl Acad. Sci. USA 96, organization of tropical deforestation in Northwest 57. Blum MJ, Bagley MJ, Walters DM, Jackson SA, 13 869–13 873. (doi:10.1073/pnas.96.24.13869) Ecuador, 1983–1995. Hum. Ecol. 26, 135–161. Daniel FB, Chaloud DJ, Cade BS. 2012 Genetic 39. Lande R, Shannon S. 1996 The role of genetic (doi:10.1023/A:1018753018631) diversity and species diversity of stream fishes variation in adaptation and population persistence 48. Arino O, Bicheron P, Achard F, Latham J, Witt R, covary across a land-use gradient. Oecologia 168, in a changing environment. Evolution 50, 434–437. Weber J-L. 2008 GlobCover: the most detailed 83–95. (doi:10.1007/s00442-011-2078-x) (doi:10.2307/2410812) portrait of Earth. ESA Bull. 136, 25–31. 58. Papadopoulou A, Anastasiou I, Spagopoulou F, 40. Mora DE, Willems P. 2012 Decadal oscillations in 49. Freitas SR, Hawbaker TJ, Metzger JP. 2010 Effects of Stalimerou M, Terzopoulou S, Legakis A, Vogler AP. rainfall and air temperature in the Paute River roads, topography, and land use on forest cover 2011 Testing the species-genetic diversity Basin–Southern Andes of Ecuador. Theor. Appl. dynamics in the Brazilian Atlantic Forest. For. correlation in the Aegean Archipelago: toward a Climatol. 108, 267–282. (doi:10.1007/s00704- Ecol. Manage. 259, 410–417. (doi:10.1016/j.foreco. haplotype-based macroecology? Am. Nat. 178, 011-0527-4) 2009.10.036) 241–255. (doi:10.1086/660828)