Biological Conservation 153 (2012) 276286

Contents lists available at SciVerse ScienceDirect

Biological Conservation
journal homepage: www.elsevier.com/locate/biocon

Identifying priority areas for island endemics using genetic versus specic diversity The case of terrestrial reptiles of the Cape Verde Islands
Raquel Vasconcelos a,b,c,, Jos Carlos Brito a,b, Slvia B. Carvalho a, Salvador Carranza c, D. James Harris a,b
a CIBIO, Centro de Investigao em Biodiversidade e Recursos Genticos da Universidade do Porto, Instituto de Cincias Agrrias de Vairo, R. Padre Armando Quintas, 4485-661 Vairo, Portugal b Departamento de Biologia, Faculdade de Cincias da Universidade do Porto, Rua Campo Alegre s/n, 4169-007 Porto, Portugal c Institute of Evolutionary Biology (CSIC-UPF), Passeig Martim de la Barceloneta, 37-49, E-08003 Barcelona, Spain

a r t i c l e

i n f o

a b s t r a c t
Genetic diversity is critical for conservation of endemic populations. It enhances adaptation to rapid environmental changes and persistence over evolutionary time-scales. In small and isolated populations, such as on islands, this is even more relevant. Nevertheless, few studies regarding the establishment of protected areas (PAs) on islands have taken genetic diversity into account. The Cape Verde Islands are in a biodiversity hotspot and present to resource planners unique problems and possibilities, hence are a good case study. This work primarily aims to compare targeting evolutionary signicant units (ESUs) versus species in reserve selection algorithms for the conservation of the endemic Cape Verdean reptile diversity by assessing the PAs network adequacy, identifying its gaps, and optimizing it based on realistic (considering areas inside PAs with lower cost) and ideal (considering all non-humanized areas with higher potential for conservation) cost scenarios. Results clearly indicate that analyses targeting ESUs are more effective in the protection of genetic diversity and less costly in terms of selected area, in total and inside PAs. Results also indicate that most ESUs and species are insufciently protected and that extra PAs are needed on most islands to reach conservation targets. Surprisingly, the total area selected in ideal and realistic prioritization scenarios are identical on most islands both for analyses targeting ESUs or species. Therefore the realistic scenario should be largely followed. The work provides an innovative methodological framework for supporting the use of genetic diversity in reserve design and its results should assist in local-scale conservation planning. 2012 Elsevier Ltd. All rights reserved.

Article history: Received 2 August 2011 Received in revised form 2 April 2012 Accepted 19 April 2012 Available online 5 July 2012 Keywords: Chioninia Ecological niche-based models ESUs versus species Hemidactylus Protected areas Tarentola

1. Introduction Genetic diversity is critical for conservation of endemic populations since it provides the raw material for the persistence of species over evolutionary time-scales, and is also of particular relevance at present time-scale in terms of providing the basis for adaptation to rapid environmental changes (Hglund, 2009). Genetic diversity is correlated with adaptive capacity of populations and tness (Soul, 1986). Furthermore, in small isolated populations, the synergy of genetic and demographic factors substantially increases their probability of extinction (Frankham, 1997). The dynamics of isolated populations can often be observed on islands (e.g. Caujap-Castells et al., 2010), which are frequently affected by catastrophic events, such as volcanic activity or droughts that can cause bottleneck effects (Whittaker and Fernn Corresponding author at: CIBIO-UP, Instituto de Cincias Agrrias de Vairo, R. Padre Armando Quintas, 4485-661 Vairo, Portugal. Tel.: +351 252 660 414; fax: +351 252 661 780. E-mail address: raquel.vasconcelos@cibio.up.pt (R. Vasconcelos).
0006-3207/$ - see front matter 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biocon.2012.04.020

dez-Palacios, 2007). Moreover, islands usually have higher numbers of endemic species than equivalent continental areas (Whittaker and Fernndez-Palacios, 2007) and high levels of uniqueness of genetic variation, especially on large or highly remote ones (Wilson et al., 2009). As a result, study and protection of endemic island taxa and their genetic diversity, considering all evolutionarily signicant units (ESUs), is particularly relevant. Designation of protected areas (PAs) safeguards habitats important to wildlife and preserves genetic resources and species diversity, provides a baseline against which human-caused changes can be measured, and allows evolutionary processes to continue without human disturbance (Quigg, 1978). The best way to represent genetic diversity in a subset of populations is to base conservation decisions on known levels of diversity within, and distribution of diversity among, populations (Neel and Cummings, 2003). Nevertheless, most studies focus on optimizing biodiversity representation at species and/or habitat level (e.g. Cowling and Pressey, 2001; Cowling et al., 2003; Bonn and Gaston, 2005; Kremen et al., 2008), while studies accounting for intra-specic genetic variability in terrestrial systems are scarce (e.g. Wei and Leberg, 2002; OMeally

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

277

and Colgan, 2005; Rissler et al., 2006; Grivet et al., 2008; Thomassen et al., 2010). To our knowledge, only four studies of this nature have been performed on island-like system. Smith et al. (2000) and Kahindo et al. (2007) studied mitochondrial lineages of avian species in African mountain regions, and considered distinctive lineages worthy of conservation. Moritz (2002) studied the conservation value of intraspecic mitochondrial lineages of rainforest fauna from northeast Australian mountains, and Setiadi et al. (2009) tested whether the two disjunct blocks constituting a National Park of an Indonesian island adequately captured the full breadth of genetic diversity of endemic species of herpetofauna. These works showed that the study of the distribution of genetic variation within species can provide useful information for biodiversity conservation. However, its concrete application in demonstrating how the selection of priority areas would perform comparatively to protecting species or habitats remains unexplored. Because nancial resources for conservation are limited, systematic methodologies and optimization algorithms have been developed to optimize biodiversity representation and persistence within PAs (Moilanen et al., 2009). The establishment of PAs is usually constrained by the existing reserve system (Pressey, 1994) and forms of land use that are, in the short term, nancially more viable than conservation (Ferrier et al., 2000). Implementation of new PAs in most developed countries is also usually hampered by high densities of human population and infrastructures. Cape Verde is an exception to some of those points, since most islands of the archipelago have less than 75 habitants/km2 (Lobban and Saucier, 2007) and few impacting human infrastructures. The PAs network in the country is sparse and was chosen in a non-systematic way, based on ad hoc presences of endemic ora and nesting bird sites and also on scenic and recreational reasons (Anonymous, 2003). Currently, only four of the 46 legally proposed terrestrial PAs of the network (Anonymous, 2003) are lawfully established and only three of them have management programs, and can thus be considered fully operational (Fig. 1), which correspond to merely 2.47% of the area of the country (IUCN and UNEP-WCMC, 2011). Additionally, biodiversity inventories are still scarce, and distribution data are still poorly documented. Hence, the degree that fully operational PAs and the ones to be implemented serve to protect important elements of biodiversity at different scales, species and genetic diversity, and the gaps in its representation are insufciently unknown. Thus, there is still a real window of opportunity to enhance the remaining 42 PAs which are not currently implemented for representing and ensuring long-term persistence of endemic biodiversity. The Cape Verde archipelago is a biodiversity-rich area, included in the Mediterranean biodiversity hotspot (Conservation International, 2005). Among vertebrates, reptiles biodiversity in the country stands out in total number of taxa and high level of endemism, since it is the richest of all Macaronesian archipelagos (Vasconcelos et al., in press). Contrary to other groups, all native taxa are endemic (Schleich, 1987) and have recently updated taxonomy, wellknown genetic diversity and dened ESUs for conservation (Arnold et al., 2008; Vasconcelos et al., 2010, 2012; Miralles et al., 2010), but were neglected in the PAs network design due to lack of distribution data. The group presents a manageable number of extant taxa, 30, within only three genera: the Hemidactylus and Tarentola geckos and the Chioninia skinks (Vasconcelos et al., in press). Hence, Cape Verdean endemic reptiles are ideal models to study reserve design and perform gap analyses taking into account genetic diversity. The aim of this study is to compare conservation planning analysis targeting ESUs or species as conservation features, with the following specic objectives: (1) to assess the adequacy of the PAs network by quantifying the protection that it guarantees or

will guarantee and its gaps, and the amount that is still missing for achieving conservation targets; (2) to map optimized priority planning units (PUs), by using two different reserve design cost scenarios: a realistic (considering PUs inside PAs with lower cost) and an ideal scenario (considering all non-humanized PUs with higher potential for conservation); (3) to evaluate the differences in the prioritization exercises, by quantifying the amount of selected PUs in total, and that will be inside PAs in each island, by assessing the percentage of target achievements for each conservation feature, and by spatially depicting the combined outputs. This work provides an innovative methodological framework for testing the usefulness of targeting genetic diversity in reserve design and its results contribute for local scale conservation planning of endemic biodiversity on islands. 2. Materials and methods 2.1. Study area Cape Verde is located in the Atlantic Ocean around 500 km off the west coast of Africa (Fig. 1). With an area of 4067 km2, the study area was divided into 76,414 grid cells, of 225 225 m each, hereafter referred as planning units (PUs), the units for reserve design. Data on the existing PAs was compiled from MAAP-DGA (2012) website. Digital maps of the terrestrial portions of the PAs network were created based on information available from government internal reports (Fig.1; Appendix A). 2.2. Taxon occurrence data and distribution models Given that only a small fraction of the territory was sampled (around 11%), and that sampled locations were spatially biased, it is most appropriate to use ecological models to predict potential distributions of occurrence, when attempting to identify priority areas for conservation (Carvalho et al., 2010). 2.2.1. Taxon occurrence data A total of 953 observations of all 30 extant Cape Verdean reptile taxa from the most recent distribution atlas were used to develop models (Vasconcelos et al., in press). For 752 observations, the geographic location was recorded with a Global Positioning System (GPS) on the WGS84 datum, whereas the remaining 201 observations were georeferenced using topographical maps to a precision of 225 m. Given that there was spatial bias in survey effort that resulted in presence clumps, observations were removed from clusters of occurrences to decrease the level of spatial autocorrelation in taxon presences (for details see Brito et al., 2009). The Nearest Neighbor Index was used to assess the degree of data clustering (clustered if < 1; dispersed if > 1): 0.42, 0.66, 0.85 and 0.89 in Chioninia delalandii, Hemidactylus boavistensis, Tarentola fogoensis and Tarentola substituta, respectively, and above 0.90 for the remaining taxa, indicating some degree of clustering for the former four species and dispersed distribution for the remaining ones. Spatial analyses were accomplished with Spatial Analyst extension of ArcGIS 9.3 (ESRI, 2008). From the available observations, 791 were used for developing distribution models for each taxon (for details see Appendix B). 2.2.2. Environmental factors Fourteen ecogeographical variables (hereafter EGVs), were used in the ecological models (Appendix C) and included altitude (Jarvis et al., 2006), slope derived from altitude with the Slope function of ArcGIS, normalized difference vegetation index (NDVI), and 11 habitat types digitized from agro-ecological and vegetation zoning

278

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

Fig. 1. Location of the study area and distribution of the protected areas (PAs) in the Cape Verde Islands (see Appendix 1 for PA designations).

maps (for details on habitats see Vasconcelos et al., 2010). NDVI 16-day L3 Global 250 m data series from 01.01.2006 to 31.12.2008 were downloaded from USGS (2009) website, corresponding to the years when sampling was performed, and then the maximum of that data series was calculated to input into the models. The Euclidean distance of each grid cell to the closest habitat-type was calculated for each individual habitat grid using the Euclidean Distance tool of ArcGIS. Advantages of using Euclidean distances to analyze habitat data include use of more than linear or point habitat features, absence of explicit error handling, and extraction of more information from data than classication-based approaches alone (Conner et al., 2003). Finally, the EGVs resolution was decreased from 0.00083 to a grid cell size average of 0.00211 (about 225 m) to match the observations resolution. 2.2.3. Predicted occurrences Models were developed for each taxon (Appendices D and E) using the Maximum Entropy approach (Phillips et al., 2006, Phillips and Dudk, 2008). This modeling technique requires only presence data as input, but consistently performs well in comparison to other methods (Elith et al., 2006), especially with low samples sizes (Wisz et al., 2008). Even so, for seven taxa with extremely low sample size (n 6 5) models were not developed. In these cases, the pixels of taxon occurrence and/or all pixels of the islet where the taxon occurs were used in subsequent analyses (Appendix B). Reptile observations and EGVs were imported into MaxEnt 3.3 software (Phillips et al., 2006). A total of 10 model replicates were run with random seed which allows a different random training/ testing data partition in each run. Observations for each replicate

were chosen by bootstrapping. Percentages assigned for testing models varied according to sample size: 10% for four taxa with less than 20 observations, 20% for 18 taxa with more than 20 observations, and 15% for one taxon with only seven observations (Appendix B). Models were run with auto-features (Phillips et al., 2006), and the Area under the Curve (AUC) of the receiver-operating characteristics (ROC) plot was taken as a measure of individual model t (Fielding and Bell, 1997). The individual model replicates (n = 10) were used to generate an average probability forecast of species occurrence (Marmion et al., 2009). Standard deviation between individual model probabilities of presence was used as an indication of prediction uncertainty (Buisson et al., 2010). Average models were reclassied to display areas of probable absence and presence for each taxon. The 10th percentile training thresholds calculated by MaxEnt were used, which considered the value of the 10 percentile species presence record to dene all areas with a lower predicted value as absent, and with a higher value as present (Phillips and Dudk, 2008). This threshold was chosen because true absence data was not available to estimate adjusted thresholds for each taxon, which would allow optimizing omission and commission rates (Liu et al., 2005). Also, it provides more parsimonious models than the minimum training presence threshold, minimizing over-predictions. To evaluate the models quality, the total observations (n = 953) were intersected with the threshold models to calculate the percentage of correct classication of presences for each taxon (Appendix B). Predicted occurrences of species that included more than one subspecies were obtained by combining the individual-taxa mod-

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

279

els using the Combine and Reclassify tools of ArcGIS. For nonmodeled taxa, predicted occurrences corresponded only to the pixels were species were detected. 2.3. Conservation planning A systematic approach was performed to identify priority PUs for the conservation of Cape Verdean endemic reptiles. To include genetic diversity into reserve design, conservation targets were rst applied on ESUs. Then, the analysis was repeated targeting species to evaluate differences in the selection of priority areas, efciency and protection of diversity. 2.3.1. Evolutionarily signicant units Considering the denition of Fraser and Bernatchez (2001) ESUs, the units for conservation action, are dened as lineages demonstrating highly restricted gene ow from other such lineages within the higher organizational level of species. Delimitation of ESUs for endemic species of each genus, based on independent mitochondrial DNA networks and signicant Hudsons Snn values, were performed on recently published papers that updated the taxonomy of the three reptile groups based on molecular markers, population and morphological analyses (see Arnold et al., 2008; Vasconcelos et al., 2010; Miralles et al., 2010). Hence, the 23 reclassied models with predicted taxon occurrences (Appendices D and E) were clipped into individual les to correspond to the 31 previously identied extant ESUs. For example, the reclassied model for C. delalandii was clipped by Santiago, Fogo, Brava and Rombos shape les, respectively, to obtain the predicted distributions of the four genetically identied lineages (see Table 1). In the case of two taxa (T. darwini and Chioninia spinalis santiagoensis) with two ESUs occurring on the same island (Santiago), the distribution data of lineages was plotted over the reclassied models to dene the extent of occurrence of each ESU and then models were clipped into two adjacent polygons. For the seven taxa for which distribution models were not developed, only observed records were accounted for reserve selection. The ESU corresponding to the extinct Chioninia coctei was not considered in the analyses, thus 38 ESUs were considered in total. 2.3.2. Adequacy of the protected areas network A gap analysis was performed to evaluate the adequacy of the PAs network. The predictive models were intersected with PAs polygons to assess the percentages of each ESU/species distribution which was currently protected or that will be protected if the full PAs network is implemented, and the amount of PUs still missing to reach conservation targets (see below). The adequacy of the PAs network was also indirectly evaluated by comparing solutions of reserve design algorithms using scenarios unconstrained or constrained to PAs (for details see Sections 2.3.3. and 2.3.4.). 2.3.3. Optimized priority planning units A software for spatial conservation prioritization, ZONATION v2.0 (Moilanen et al., 2009) was used to evaluate if the PAs network was optimal for protecting all ESUs of endemic reptiles from Cape Verde and to compare effectiveness of two analysis, one targeting ESUs versus another targeting species as conservation features. ZONATION uses a gradient-like iterative heuristic, which gives a solution very close to the global optimal (van Teeffelen and Moilanen, 2008) to produce a sequential removal of PUs throughout the planning region. PUs with less conservation value are removed rst, thus, PUs with highest rank have highest conservation value. Target-based planning was chosen as PUs removing rule since the goal was to nd the best solution in which the maximum number of ESUs or species met conservation targets. Conservation targets were set as 12% for this analysis because it appears to be

widely used in similar analyses (Wright and Mattson, 1996; Cant et al., 2004). Hence, any resource category with at least 12% of its area protected was considered adequately protected. The only exception was applied to taxa considered endangered (Critically Endangered or Endangered), according to IUCN Red List criteria (Vasconcelos et al., in press) to which a higher target was set (100%), following recommendations of similar works (Carvalho et al., 2010; Jackson et al., 2004). In order to generate spatial aggregation into the solution, the rule only remove from edges was selected. The warp factor was set to one with an aggregation level of 0.04. The Boundary Length Penalty (BLP), which devalues reserve structures with lots of edge, was chosen as the method for inducing reserve network aggregation (Moilanen and Kujala, 2008). Two cost scenarios, one realistic and one ideal, were simulated in each analysis, constrained and unconstrained by the 46 PAs, respectively. In realistic scenarios, cells with main roads and small and large urban areas and other infrastructures (with a buffer radius of 112 m or 1 km, respectively) were given a cost of 100, with secondary roads a cost of 75, with PAs 1, and remaining cells 50. All different PA categories were thus treated with the same weight. In ideal scenarios, PAs were not taken into account, thus cells with main roads and urban areas were given a cost of 100, with secondary roads 50, and remaining cells 1. The minimum set of PUs with higher rank in the nal solution, which assured that all ESUs or species were represented with the desired target, was selected for each scenario in reclassied binary les. 2.3.4. Differences in the prioritization exercises The selected PUs in each scenario (ideal or realistic) and analyses targeting different conservation features (ESUs or species) were counted and intersected with PAs polygons using ArcGIS to calculate the amount of PUs encompassed in the PAs network, and to identify differences in outputs according to cost scenarios and target analyses in total and per island/islet. The selected PUs in each cost scenario and target analyses were also intersected with predicted occurrences to estimate the percentages of conservation targets achievement of each ESU and species. The output calculations of the two analyses (targeting ESUs or species) were subtracted to quantify differences in amount and percentage of selected PUs between the two prioritization exercises per island/islet and per conservation feature. To identify where PUs selected by realistic scenarios differed from ideal ones, these were combined for each analysis using ArcGIS. To identify areas where the analysis targeting ESUs differed from the one targeting species, the analogous scenarios of each analysis were combined using ArcGIS. 3. Results 3.1. Evaluation of ecological niche-based models The ROC plots exhibited high average AUCs with low standard deviations (SD) for both training and test datasets in all model types (Appendix B). Average AUCs for training and test datasets were 0.985 0.003 and 0.970 0.018, respectively. Thresholded models identied suitable cells for each species. The average percentage of observations identied in suitable cells was 80.8% (Appendix B). 3.2. Adequacy of the protected areas network Presently, with three PAs fully operational, only Chioninia spinalis spinalis fulls the conservation target (Table 1). All remaining ESUs are insufciently protected. When considering all PAs to be implemented, these gures are quite different. In these

280

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

Table 1 Number (n) of planning units (PUs) where each evolutionarily signicant unit (ESU) and species are predicted to occur, targeted for conservation, inside protected areas (PAs) fully operational (present) or to be implemented (future), and missing to meet conservation targets (see Section 2 for details). Percentages (%) are given between brackets. Taxon/ESU ESU Species Island/Islet Predicted Targeted Inside PAs Present n Hemidactylus lopezjuradoi H. boavistensis ESU Sal ESU Boavista H. bouvieri H. bouvieri, SN population H. bouvieri bouvieri H. bouvieri razoensis Tarentola boavistensis T. bocagei T. fogoensis T. darwini ESU North ESU South T. substituta T. raziana T. caboverdiana T. nicolauensis T. gigas T. gigas brancoensis T. gigas gigas T. rudis T. maioensis T. protogigas T. protogigas protogigas T. protogigas hartogi ESU Brava ESU Rombos Chioninia vaillanti C. vaillanti vaillanti C. vaillanti xanthotis C. delalandii ESU Santiago ESU Fogo ESU Brava ESU Rombos C. nicolauensis C. fogoensis C. stangeri ESU Desertas ESU S. Vicente C. spinalis C. spinalis salensis C. spinalis santiagoensis ESU North ESU South C. spinalis spinalis C. spinalis maioensis C. spinalis boavistensis Total                                           38 21   F BV, S S BV SN, SA, SL, ra SN SA SL, ra BV SN F ST ST ST SV SL, ra, br SA SN br, ra br ra ST M F, B, ro F B, ro B ro ST, F, ro ST F, ro ST, F, B, ro ST F B ro SN SA SV, SL, ra, br SL, ra, br SV S, ST, F, M, BV S ST ST ST F M BV 10 + 3 1 5889 2225 3664 111 2 1 108 2994 384 1099 9801 3819 5982 1934 582 4180 2359 153 46 107 2380 2013 671 4 608 59 4084 3510 574 7828 4541 2238 990 59 1432 3668 1046 811 235 16,345 2356 684 4056 2118 1635 5496 68,964 n 1 707 267 440 111 2 1 108 359 46 132 1176 458 718 232 71 502 283 153 46 107 286 242 671 4 73 7 4084 3510 574 939 545 269 119 7 172 440 1046 811 235 1961 283 82 487 254 196 660 8276 % (100) (12) (12) (12) (100) (100) (100) (100) (12) (12) (12) (12) (12) (12) (12) (12) (12) (12) (100) (100) (100) (12) (12) (100) (100) (12) (12) (100) (100) (100) (12) (12) (12) (12) (12) (12) (12) (100) (100) (100) (12) (12) (12) (12) (12) (12) (12) (12) n 0 0 0 0 1 1 0 0 0 0 17 153 153 0 0 0 0 78 0 0 0 0 0 0 0 0 0 157 157 0 253 167 86 0 0 149 0 0 0 0 303 0 0 0 303 0 0 1111 % (0.0) (0.0) (0.0) (0.0) (0.9) (50.0) (0.0) (0.0) (0.0) (0.0) (1.5) (1.6) (4.0) (0.0) (0.0) (0.0) (0.0) (3.3) (0.0) (0.0) (0.0) (0.0) (0.0) (0.0) (0.0) (0.0) (0.0) (3.8) (4.5) (0.0) (3.2) (3.7) (3.8) (0.0) (0.0) (10.4) (0.0) (0.0) (0.0) (0.0) (1.4) (0.0) (0.0) (0.0) (14.3) (0.0) (0.0) (1.6) Future n 0 2986 517 2469 109 1 0 108 1261 0 17 318 153 165 24 582 97 78 153 46 107 29 601 59 0 0 59 292 233 59 483 338 86 0 59 149 319 811 811 0 4274 465 0 0 303 559 2947 12,652 % (0.0) (50.7) (23.2) (67.4) (98.2) (50.0) (0.0) (100.0) (42.1) (0.0) (1.5) (3.2) (4.0) (2.8) (1.2) (100.0) (2.3) (3.3) (100.0) (100.0) (100.0) (1.2) (29.9) (8.8) (0.0) (0.0) (100.0) (7.1) (6.6) (10.3) (6.2) (7.4) (3.8) (0.0) (100.0) (10.4) (8.7) (77.5) (100.0) (0.0) (26.1) (19.7) (0.0) (0.0) (14.3) (34.2) (53.6) (18.3) n 1 0 0 0 2 1 1 0 0 46 115 858 305 553 208 0 405 205 0 0 0 257 0 612 4 73 0 3792 3277 515 456 207 183 119 0 23 121 235 0 235 0 0 82 487 0 0 0 4376 % (100.0) (0.0) (0.0) (0.0) (1.8) (50.0) (100.0) (0.0) (0.0) (100.0) (87.1) (73.0) (66.6) (77.0) (89.7) (0.0) (80.7) (72.4) (0.0) (0.0) (0.0) (89.8) (0.0) (91.2) (100.0) (100.0) (0.0) (92.9) (93.4) (89.7) (48.6) (38.0) (68.0) (100.0) (0.0) (13.3) (27.5) (22.5) (0.0) (100.0) (0.0) (0.0) (100.0) (100.0) (0.0) (0.0) (0.0) (52.9) Missing

   

    

  

  

Islands/Islets: SA, Santo Anto; SV; S. Vicente; SL, Santa Luzia; br, Branco; ra, Raso; SN, S. Nicolau; S, Sal; BV, Boavista; M, Maio; ST, Santiago; F, Fogo; B, Brava; ro, Rombos.

circumstances, 15 of the 38 ESUs potential distributions (40%) considered in the analyses will have the targeted percentage of its distribution inside PAs (Table 1). However, 9 of those 38 ESUs (24%) would not have a single PU inside a PA and only Hemidactylus bouvieri razoensis, Tarentola gigas brancoensis and T. g. gigas would be fully protected (Table 1). If conservation priorities targeted species instead of ESUs, the present PAs would fail to achieve conservation targets for all of them, and after the establishment of the full PAs network two of them would not have a single PU inside a PA (10%), and only six out of 21 species (29%) would reach conservation targets (Table 1). The PAs network adequacy was also indirectly evaluated by comparing the ideal and realistic cost scenarios, unconstrained and constrained to the PAs network, respectively (see Section 3.4).

3.3. Optimized priority planning units The maps with selected priority PUs for conservation, considering realistic and ideal scenarios of analyses considering ESUs or species as conservation targets are presented in Fig. 2A and B, respectively.

3.4. Differences in the prioritization exercises Overall, in both analyses targeting ESUs or species, more PUs in total and inside PAs were selected in the realistic than in the ideal scenarios (8127/8558 versus 8036/8458 total PUs and 30.9/32.8% versus 21.2/21.9% of PUs inside PAs) (Table 2).

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

281

Fig. 2. Selected planning units (PUs) necessary to reach conservation targets for reptile ESUs (A) and species (B) from Cape Verde Islands considering the realistic and ideal scenarios inside and outside protected areas (see Section 2 and Appendices FJ for details).

In total, more PUs in total (5% on average) and lower percentages of PUs inside PAs (up to 7% less) were selected in the analysis targeting species than ESUs, comparing analogous scenarios (Table 2). Additionally, more ESUs and species reach conservation targets more efciently in the analysis targeting ESUs (Table 3). Apart from quantitative differences, the spatial conguration of the selected PUs differed considerably between the two target analyses in some islands (see Section 3.4.1 and Fig. 3). 3.4.1. Differences per island/islet Variation in the amount of PUs (in total and within PAs) selected within each island/islet was large. For four islands and three islets, almost the same amount of PUs was selected, both in total and within PAs, in both cost scenarios and target analyses (Santo Anto, S. Vicente, Sta. Luzia, Branco, Raso, S. Nicolau and Rombos) while for other islands a large difference was detected between solutions. When just comparing results obtained with ideal and realistic cost scenarios, differences between the total number of selected PUs was very low for most islands, both when targeting ESUs or species. Using ESUs as target units, solutions were almost identical in all islands, except on Sal, where the ideal model would be more effective, since much less area would be needed to protect reptile diversity (Table 2). In this analysis, it was also on Sal where PUs selected by each scenario spatially coincided by the least amount, followed by Maio; in the remaining islands, concordance of PUs selected by both scenarios was relatively high (Fig. 2A and Appendices FJ). Using species as target units, solutions only substantially differed on Boavista, Santiago, Fogo and Maio islands and only on Boavista did the ideal scenario perform better than the realistic one (Table 2). In these cases, it is again on these islands, especially Maio and Boavista, where the PUs selected by each scenario spatially coincided by the least amount (Fig. 2B). However, as expected, differences between cost scenarios were higher when considering PUs selected inside PAs, which were considerably larger in the realistic scenario in several islands regardless of targeting ESUs or species, except on the uninhabited islands/islets of the archipelago (Rombos and Desertas group), where 100% of selected PUs would be inside PAs regardless of the cost scenario and targeted feature for conservation (Table 2). Comparing just targeted conservation features, they little differed in the above referred four islands and islets and presented

almost the same percentage of selected PUs inside PAs on ve others (Santiago, Fogo, Brava, Boavista and Maio) in a least one cost scenario. However, they differed in the number of total selected PUs on six islands (Sal, Santiago, Boavista, Brava, Maio and Fogo), with the analyses targeting species selecting no PUs on Sal and more total PUs than the analysis targeting EUSs in the latter four islands. There were also noticeable differences in the number of selected PUs inside PAs. In the analysis using ESUs as targets, Sal would reach 100% of PUs inside PAs in the realistic scenario, while Fogo and Brava would present all PUs selected by both scenarios unprotected (Table 2 and Fig. 2A). On the contrary, in the analyses using species as targets, Sal would not have a single PU inside PAs in both scenarios, while some PUs would be selected inside PAs on Fogo, under the realistic scenario, although below the 12% threshold of protection (Table 2 and Fig. 2B). Spatial agreement between analyses targeting species or ESUs was relatively high in most islands. However, on Sal analyses totally differ and on many northern and southern Santiago cells where PUs were selected only on the analyses targeting ESUs, and also in most area of Brava, where most PUs are only prioritized by the analysis targeting species (Fig. 3). 3.4.2. Differences per conservation feature (ESUs or species) Considering ESUs, 19 out of the 38 presented no differences in the amount of selected PUs to meet conservation targets between analyses targeting different conservation features in both cost scenarios (Table 3). However, the analysis targeting species was more costly for 10 ESUs, and considerably more costly (selecting 50% more PUs) in at least one cost scenario for eight of them. In addition to that, it failed to reach conservation targets for four ESUs (H. boavistensis and C. s. salensis from Sal and northern and southern ESUs of C. s. santiagoensis from Santiago), completely failing targets for three of them in both cost scenarios (Table 3). The analysis targeting species was slightly more effective only for four ESUs considering at least one cost scenarios, and considerably better for one (Tarentola rudis), considering the realistic scenario only (Table 3). Considering species, 12 out of the 21 presented no differences between analyses targeting different conservation features in the amount of selected PUs to meet conservation targets in both cost scenarios (Table 3). However, the analysis targeting species was

282 Table 2 Number (n) of total planning units (PUs) and PUs inside the 46 protected areas (PAs) on each island/islet available and selected by each model scenario and target analysis, and subtracted between the two target analyses. Percentages (%) are given between brackets.

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

Realistic scenario

0 1 0 0 0 0 374 518 112 323 15 482 0

33.2

Total

Species ESUs target analyses

(0.0) (0.0) (0.0) (0.0) (0.0) (0.6) (21.6) (1.9) (3.3) (0.3) (0.0) (0.0) (0.0)

(1.3)

(0.7)

431

less effective for six species and considerably less effective in at least one cost scenario for ve of them. Targeting species was therefore slightly more effective only for three species, and considerably better for one (Table 3). In summary, solutions found when targeting species failed to achieve targets for some ESUs and were generally more costly, but the inverse was not true, as solutions found when targeting ESUs successfully achieved targets for all species and more efciently, regardless of the cost scenario used.

(0.2) (0.0) (0.0) (0.0) (0.0) (0.0) (100.0) (0.0) (0.0) (1.1) (2.8) (0.0) (0.0)

Inside PAs

1 0 0 0 0 0 374 518 112 25 15 0 0

22.8

(7.4)

297

(1.9)

4. Discussion 4.1. Evaluation of the methodology

Inside PAs

0 0 0 0 0 2 61 149 59 0 0 0 0

11.2

Ideal scenario

145

Ecological niche-based models provided fairly robust predictions of occurrences and the reserve design algorithm identied priority PUs for conservation of endemic reptiles. Hence, this study may turn into an important tool in planning and designation of PAs in Cape Verde. Additionally, the novel approach used may prove useful to other studies attempting to maximize representation of genetic diversity in conservation prioritization. It also showed the increased importance of targeting ESUs instead of species as conservation features. Several studies assumed predicted probabilities of occurrence of taxa to be surrogates of probability of persistence, and targeted areas where probabilities were high (e.g. Margules and Stein, 1989; Williams and Arajo, 2000). In this study, the most effective selection algorithm incorporated the most probable occurrence areas of all ESUs, thus potentially enhancing taxa persistence even more. Nevertheless, potential pitfalls might have emerged because patterns of neutral variation, as measured by molecular markers, may not reect levels of adaptive variation for all traits across all populations (Le Corre and Kremer, 2003). However, given the difculty in measuring adaptive variation for wild species, molecular markers are valuable surrogates and, in some cases, may be conservative estimates of the expectations of loss and recovery of quantitative genetic variation (Lynch et al., 1999). Additionally, adaptive features may be best protected by maintaining the context for selection, such as heterogeneous landscapes (Hglund, 2009). Since habitats in this archipelago are most different among islands than within them (data not shown), targeting 12% of the area of each island for conservation potentially enhances the cover of adaptive variation. Another question that might be addressed is if ecological models should have been based on ESUs instead of taxa. Nevertheless, in doing so, sample sizes would have been greatly reduced, which would probably compromise analytical methods chosen. More importantly, the comparison of targeting ESUs instead of species would be biased, as some predictions of occurrence would be necessarily different from taxa predictions, as in the case of species with no subspecies occurring on different islands and subspecies with more than one ESU. Furthermore, in those cases, the probability of occurrence of those ESUs might be more related to isolation by the ocean, genetic drift and other factors than with ecogeographical factors, as frequently they are not reciprocally monophyletic (see Arnold et al., 2008; Miralles et al., 2010; Vasconcelos et al., 2010), and do not present any evidence for local adaptation, a determinant condition for modeling them (Pearman et al., 2010). Concerning the representation targets set for each ESU and species, choosing the 12% target for non-endangered taxa makes these results comparable to other natural resources conservation studies but does not suggest that this gure has any established scientic validity to assure that populations selected for conservation are viable. The question of which target should be used is a paramount in conservation planning but remains largely unsolved (Tear et al.,

0 0 0 0 0 0 283 198 126 189 88 482 0

32.5 216 658 (39.9) 142 651 193 625 (41.2) 131 618 (39.7) 5897 Average 831 (56.9) (49.5)

Total

(14.5) (4.5) (100.0) (100.0) (100.0) (15.6) (0.0) (100.0) (100.0) (6.6) (2.8) (0.0) (100.0)

Inside PAs

Realistic scenario

73 19 655 49 108 52 0 1173 356 250 15 0 59

505 420 655 49 108 334 0 1173 356 3762 536 601 59

Total

Species target analysis

Inside PAs

Ideal scenario

7 0 655 49 108 9 0 589 149 225 0 0 59

(1.4) (0.0) (100.0) (100.0) (100.0) (2.7) (0.0) (68.9) (40.5) (5.8) (0.0) (0.0) (100.0)

505 420 655 49 108 334 0 855 368 3895 609 601 59

Total

Inside PAs

Realistic scenario

72 19 655 49 108 52 374 655 244 225 0 0 59

(14.3) (4.5) (100.0) (100.0) (100.0) (15.6) (100.0) (100.0) (100.0) (5.5) (0.0) (0.0) (100.0)

505 419 655 49 108 334 374 655 244 4085 521 119 59

Total

ESUs target analysis

Inside PAs

Ideal scenario

7 0 655 49 108 11 61 440 90 225 0 0 59

(1.4) (0.0) (100.0) (100.0) (100.0) (3.3) (21.6) (67.0) (37.2) (5.5) (0.0) (0.0) (100.0)

505 420 655 49 108 334 283 657 242 4084 521 119 59

Total

Inside PAs

(4.0) (1.4) (100.0) (100.0) (100.0) (4.0) (20.6) (39.8) (25.0) (2.2) (19.4) (0.0) (100.0)

Available

14,899 4275 656 49 108 6515 4187 11930 5141 18829 8841 1176 59

Total

Santo Anto S. Vicente Santa Luzia Branco Raso S. Nicolau Sal Boavista Maio Santiago Fogo Brava Rombos

Island/Islet

Total

76665

10799

601 58 656 49 108 258 862 4743 1283 409 1713 0 59

(14.1)

8036

1705

(21.2)

8127

2512

(30.9)

8458

1850

(21.9)

8558

2809

(32.8)

422

Table 3 Number (n) of planning units (PUs) where each evolutionarily signicant unit (ESU) and species are predicted to occur, targeted for conservation, selected considering each target analyses and cost scenarios to meet conservation targets (in bold if higher than 150% and in italic if not reaching targets), and subtracted between the two target analyses (see Section 2 for details). Percentages (%) are given between brackets. Taxon/ESU Island/islet ESU Species Predicted Targeted ESUs target analysis Ideal scenario n Hemidactylus lopezjuradoi H. boavistensis ESU Sal ESU Boavista H. bouvieri H. bouvieri, SN population H. bouvieri bouvieri H. bouvieri razoensis Tarentola boavistensis T. bocagei T. fogoensis T. darwini ESU North ESU South T. substituta T. raziana T. caboverdiana T. nicolauensis T. gigas T. gigas brancoensis T. gigas gigas T. rudis T. maioensis T. protogigas T. protogigas protogigas T. protogigas hartogi ESU Brava ESU Rombos Chioninia vaillanti C. vaillanti vaillanti C. vaillanti xanthotis C. delalandii ESU Santiago ESU Fogo ESU Brava ESU Rombos C. nicolauensis C. fogoensis C. stangeri ESU Desertas ESU S. Vicente C. spinalis C. spinalis salensis C. spinalis santiagoensis ESU North ESU South C. spinalis spinalis C. spinalis maioensis C. spinalis boavistensis F BV, S S BV SN, SA, SL, ra SN SA SL, ra BV SN F ST ST ST SV SL, ra, br SA SN br, ra br ra ST M F, B, ro F B, ro B ro ST, F, ro ST F, ro ST, F, B, ro ST F B ro SN SA SV, SL, ra, br SL, ra, br SV S, ST, F, M, BV S ST ST ST F M BV                                             1 5889 2225 3664 111 2 1 108 2994 384 1099 9801 3819 5982 1934 582 4180 2359 153 46 107 2380 2013 671 4 608 59 4084 3510 574 7828 4541 2238 990 59 1432 3668 1046 811 235 16,345 2356 684 4056 2118 1635 5496 n 1 707 267 440 111 2 1 108 359 46 132 1176 458 718 232 71 502 283 153 46 107 286 242 671 4 73 7 4084 3510 574 939 545 269 119 7 172 440 1046 811 235 1961 283 82 487 254 196 660 % (100) (12) (12) (12) (100) (100) (100) (100) (12) (12) (12) (12) (12) (12) (12) (12) (12) (12) (100) (100) (100) (12) (12) (100) (100) (12) (12) (100) (100) (100) (12) (12) (12) (12) (12) (12) (12) (100) (100) (100) (12) (12) (12) (12) (12) (12) (12) n 1 940 283 657 111 2 1 108 652 47 220 2936 1343 1593 233 582 502 284 153 46 107 529 242 330 4 104 59 4084 3510 574 2751 2248 325 119 59 285 504 1046 811 235 2595 283 83 487 270 242 660 % (100) (133) (106) (149) (100) (100) (100) (100) (182) (102) (167) (250) (293) (222) (100) (820) (100) (100) (100) (100) (100) (185) (100) (49) (100) (142) (843) (100) (100) (100) (293) (412) (121) (100) (843) (166) (115) (100) (100) (100) (132) (100) (101) (100) (106) (123) (100) Realistic scenario n 1 922 267 655 111 2 1 108 652 47 220 2908 1339 1569 233 582 502 284 153 46 107 515 242 330 4 104 59 4084 3510 574 2686 2183 325 119 59 283 475 1046 811 235 2597 283 83 487 270 244 660 % (100) (130) (100) (149) (100) (100) (100) (100) (182) (102) (167) (247) (292) (219) (100) (820) (100) (100) (100) (100) (100) (180) (100) (49) (100) (142) (843) (100) (100) (100) (286) (401) (121) (100) (843) (165) (108) (100) (100) (100) (132) (100) (101) (100) (106) (124) (100) Species target analysis Ideal scenario n 1 855 0 855 111 2 1 108 850 47 306 2747 1256 1491 233 582 502 284 153 46 107 423 368 1338 4 608 59 4084 3510 574 3224 2162 413 590 59 285 504 1046 811 235 2341 0 0 381 358 368 853 % (100) (121) (0) (194) (100) (100) (100) (100) (237) (102) (232) (234) (274) (208) (100) (820) (100) (100) (100) (100) (100) (148) (152) (199) (100) (833) (843) (100) (100) (100) (343) (397) (154) (496) (843) (166) (115) (100) (100) (100) (119) (0) (0) (78) (141) (188) (129) Realistic scenario n 1 1048 0 1048 111 2 1 108 789 47 235 2547 1256 1291 233 582 502 284 153 46 107 286 356 1338 4 608 59 4084 3510 574 3199 2218 332 590 59 283 474 1046 811 235 2108 0 0 148 285 356 1171 % (100) (148) (0) (238) (100) (100) (100) (100) (220) (102) (178) (217) (274) (180) (100) (820) (100) (100) (100) (100) (100) (100) (147) (199) (100) (833) (843) (100) (100) (100) (341) (407) (123) (496) (843) (165) (108) (100) (100) (100) (107) (0) (0) (30) (112) (182) (177) Species ESUs target analyses Ideal scenario n 0 85 283 198 0 0 0 0 198 0 86 189 87 102 0 0 0 0 0 0 0 106 126 1008 0 504 0 0 0 0 473 86 88 471 0 0 0 0 0 0 254 283 83 106 88 126 193 % (0) (12) (106) (45) (0) (0) (0) (0) (55) (0) (65) (16) (19) (14) (0) (0) (0) (0) (0) (0) (0) (37) (52) (150) (0) (690) (0) (0) (0) (0) (50) (16) (33) (396) (0) (0) (0) (0) (0) (0) (13) (100) (101) (22) (35) (64) (29) Realistic scenario n 0 126 267 393 0 0 0 0 137 0 15 361 83 278 0 0 0 0 0 0 0 229 114 1008 0 504 0 0 0 0 513 35 7 471 0 0 1 0 0 0 489 283 83 (339) 15 112 511 % (0) (18) (100) (89) (0) (0) (0) (0) (38) (0) (11) (31) (18) (39) (0) (0) (0) (0) (0) (0) (0) (80) (47) (150) (0) (690) (0) (0) (0) (0) (55) (6) (3) (396) (0) (0) (0) (0) (0) (0) (25) (100) (101) (70) (6) (57) (77)

   

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

    

  

  

Islands/Islets: SA, Santo Anto; SV; S. Vicente; SL, Santa Luzia; br, Branco; ra, Raso; SN, S. Nicolau; S, Sal; BV, Boavista; M, Maio; ST, Santiago; F, Fogo; B, Brava; ro, Rombos. 283

284

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

Fig. 3. Combination of selected planning units (PUs) necessary to reach conservation targets for reptiles ESUs and species from Cape Verde Islands considering the ideal (A, reddish colors) and realistic (B, bluish colors) scenarios inside and outside protected areas (see Section 2 for details). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

2005), although the solution might involve using variable targets regarding taxon ranges (Rodrigues et al., 2004; Maiorano et al., 2007). Regarding cost scenarios, the selected PUs of ideal models spatially coincided with the realistic model scenarios in most cases (Fig. 2). In addition, different scenarios presented identical efciency (similar number of selected PUs) in most islands. These results are surprising since ideal scenarios theoretically minimize costs for PUs selection in comparison with scenarios constrained by PAs, because PAs are generally biased for other factors rather than protecting biodiversity. Three complementary rationales might explain this result. First, both scenarios are congruent in selecting many PUs outside PAs, in order to encompass 12% of most ESUs or species distributions (100% in threatened taxa), and in those areas, selected PUs by both scenarios are likely to overlap. Second, in regions less affected by anthropogenic disturbance, selecting an ideal network that maximizes representation of diversity most times does not lead to signicantly different results from a selection by chance (Bonn and Gaston, 2005). This pattern might be especially noticeable in small areas. In Cape Verde, where few impacting human infrastructures are present in most islands and PAs were designed using ad-hoc criteria, reptile distributions are little restricted by anthropogenic actions. Thus, some PUs inside PAs selected by the realistic scenarios (that prioritizes PUs inside PAs) are likely to be also selected by the ideal models. This is most noticeable on islands like Boavista, Maio and Santiago. Third, alternatively, the extensive overlap of solutions from different models may suggest that PUs selected ad hoc for other endemic groups on which PAs locations were based on, such as birds, are also good for reptiles and vice-versa. Thus, reptiles may be good surrogates of priority PUs for endemic birds and ora, although they might not be as good for other groups such as invertebrates (Rodrigues and Gaston, 2001). In fact, some recent ad hoc data on endemic birds conrms several selected PUs outside PAs depicted by this work as important for conservation. For instance, the threatened Cape Verde cane warbler (Acrocephalus brevipennis) also occurs on the northeast of Fogo, and a large colony of the Critically Endangered purple heron was conrmed (Ardea purpura) around Montanha, Santiago (see Appendices I and J, respectively), following Hazevoet (2010). It would be important to cross-update information about

georeferenced nesting sites of the endemic birds and accurate distribution maps of the endemic ora with the performed analyses, whenever they become available, to conrm this result. Regarding the analyses using different conservation targets, it is quantitatively more effective and less costly to target ESUs since a lower number of total PUs are selected (around 5% less) in comparison to targeting species. This is mainly because less PUs would be selected targeting ESUs on Brava, Boavista, Maio and Santiago. The selection of less PUs on Brava is explained by very different conservation status and associated conservation targets (12% and 100%) of the two subspecies of Tarentola protogigas, forcing the analysis targeting species to select much more PUs to protect the endangered species as a whole. This fact reinforces the importance of using subspecic taxonomy to evaluate conservation status and to prioritize areas for conservation (Neel and Cummings, 2003). On the other islands, that is explained by the lower number of targeted PUs per island when targeting ESUs due to the species fragmentation among different islands/islets corresponding to different ESUs. Although the analysis targeting species presented higher total numbers of selected PUs inside some PAs (around 9% more) when comparing to the analysis targeting ESUs, the overall percentage of PUs inside PAs is lower (Table 2). The rst is explained by the fact that aggregation of the selected PUs is necessarily easier when targeting fewer features (21 species versus 38 ESUs), what might seem a disadvantage of targeting ESUs, and the latter because no PUs were selected in Sal in that analyses. However, the choice of targeting species instead of ESUs would also necessarily imply failing to reach conservation targets for four unique lineages, completely for three of them (two from Sal and one from northern Santiago; see Section 3.4.2.). Moreover, it would generally imply the selection of considerably more PUs per ESU and species. Hence the analyses targeting ESUs is also qualitatively more effective for the conservation of diversity at different levels, with the additional advantage of considering its persistence. 4.2. Adequacy of the protected areas network and optimized priority planning units Currently, Cape Verde presents the lowest proportion of land (about 2%) devoted to conservation in comparison to several other

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286

285

oceanic islands (40% on average; Caujap-Castells et al., 2010). The implementation of the full PAs network is thus needed to guarantee partial protection of the biodiversity of the endemic reptiles and their habitats (Tables 2 and 3). In addition, implementation of new PAs based on the proposed scenarios targeting ESUs, is needed to fully protect the genetic diversity of these reptiles. Otherwise, nine ESUs and two species, Hemidactylus lopezjuradoi and Tarentola bocagei, both considered threatened species (following Vasconcelos et al., in press), would fail completely to reach targets and several other threatened taxa would not be adequately protected (Hemidactylus bouvieri, Tarentola boavistensis, Tarentola raziana, T. rudis, T. protogigas, Chioninia vaillanti, and Chioninia stangeri). Reserve design analyses targeting ESUs indicated two main patterns in the Cape Verde Islands. On a group of islands, namely Desertas group, Sal, Boavista, Maio and Rombos, designation of new PAs is not a priority, since PAs that are going to be implemented will guarantee total targeted protection of all endemic reptile species and ESUs occurring there and their habitats. On the remaining islands, the planned PAs are clearly insufcient, since about 60% of ESUs would not achieve conservation targets (Table 13 and Appendices FJ). In the extreme case of Fogo and Brava islands, no scenario selected a single PU inside a PA using ESUs as conservation targets (Table 2 and Appendix I). Recommended conservation actions are described in Appendix K. This study addresses one of the major constraints of conservation in the Cape Verde Islands, namely the lack of basic information in formats that policymakers and administrators can interpret and use (Miller, 1993). It is expected that this innovating framework can be applied to other island systems with well-know genetic diversity such as the Canary Islands, where extensive work has been carried out on the endemic reptiles (e.g. Brown and Pestano, 1998; Carranza et al., 2002; Cox et al., 2010) or other island-like systems, such as mountain ranges. Acknowledgments R.V. is grateful to I. Gomes from INIDA (Instituto Nacional de Investigao e Desenvolvimento Agrrio) and A. Fernandes from DGA (Direco Geral do Ambiente) for sharing the government internal reports; to Dr. J. Spencer, Major A. Rocha, and Eng. J. Andrade from Direco de Servio e Cartograa e Cadastro, for helping and facilitating digital data of roads and urban areas of Cape Verde. This study was partially supported by Fundao para a Cincia e Tecnologia (FCT): SFRH/BD/25012/2005 and SFRH/BPD/79913/ 2011 (to R.V.), PTDC/BIA-BDE/74288/2006; J.C.B. and D.J.H. have FCT contracts (Programa Cincia 2007 and 2008 Fundo Social Europeu); and Ministerio de Educacin y Ciencia CGL200911663/BOS.

Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.biocon.2012. 04.020. References
Anonymous, 2003. DL n 3/2003, 24 de Fevereiro. Boletim Ocial da Repblica de Cabo Verde de 2003. N 5, I srie. Ministrio da Justia, Praia, Cabo Verde. Arnold, E.N., Vasconcelos, R., Harris, D.J., Mateo, J.A., Carranza, S., 2008. Systematics, biogeography and evolution of the endemic Hemidactylus geckos (Reptilia, Squamata, Gekkonidae) of the Cape Verde Islands: based on morphology and mitochondrial and nuclear DNA sequences. Zool. Scr. 37, 619636. Bonn, A., Gaston, K.J., 2005. Capturing biodiversity: selecting priority areas for conservation using different criteria. Biodivers. Conserv. 14, 10831100.

Brito, J.C., Acosta, A.L., lvares, F., Cuzin, F., 2009. Biogeography and conservation of taxa from remote regions: an application of ecological-niche based models and GIS to North-African Canids. Biol. Conserv. 142, 30203029. Brown, R.P., Pestano, J., 1998. Phylogeography of skinks (Chalcides) in the Canary Islands inferred from mitochondrial DNA sequences. Mol. Ecol. 7, 11831191. Buisson, L., Thuiller, W., Casajus, N., Lek, S., Grenouillet, G., 2010. Uncertainty in ensemble forecasting of species distribution. Glob. Change Biol. 16, 11451157. Cant, C., Wright, R.G., Scott, J.M., Strand, E., 2004. Assessment of current and proposed nature reserves of Mexico based on their capacity to protect geophysical features and biodiversity. Biol. Conserv. 115, 411417. Carranza, S., Arnold, E.N., Mateo, J.A., Geniez, P., 2002. Relationships and evolution of the North African geckos, Geckonia and Tarentola (Reptilia: Gekkonidae), based on mitochondrial and nuclear DNA sequences. Mol. Phylogenet. Evol. 23, 244256. Carvalho, S.B., Brito, J.C., Pressey, R.L., Crespo, E., Possingham, H.P., 2010. Simulating the effects of using different types of species distribution data in reserve selection. Biol. Conserv. 143, 426438. Caujap-Castells, J., Tye, A., Crawford, D.J., Santos-guerra, A., Sakai, A., Beaver, K., Lobin, W., Florens, F.B.V., Moura, M., Jardim, R., Gomes, I., Kueffer, C., 2010. Conservation of oceanic island oras: present and future global challenges. Perspect. Plant Ecol. Evol. Syst. 12, 107129. Conner, L.M., Smith, M.D., Burger, L.W., 2003. A comparison of distance-based and classication-based analyses of habitat use. Ecology 84, 526531. Conservation International, 2005. Conservation International Hotspots. <http:// www.biodiversityhotspots.org/xp/Hotspots/hotspotsScience/pages/hotspots_ revisited.aspx>. Cowling, R.M., Pressey, R.L., 2001. Rapid plant diversication: planning for an evolutionary future. Proc. Nat. Acad. Sci. U.S.A. 98, 54525457. Cowling, R., Pressey, R.L., Rouget, M., Lombard, A.T., 2003. A conservation plan for a global biodiversity hotspot the Cape Floristic Region, South Africa. Biol. Conserv. 112, 191216. Cox, S.C., Carranza, S., Brown, R.P., 2010. Divergence times and colonization of the Canary Islands by Gallotia lizards. Mol. Phylogenet. Evol. 56, 747757. Elith, J., Graham, C.H., Anderson, R.P., Dudyk, M., Freer, S., Guisan, A., Hijmans, R.J., Huettmann, F., Leathwick, J.R., Lehmann, A., Li, J., Lohmann, L.G., Loiselle, B.A., Manion, G., Moritz, C., Nakamura, M., Nakazawa, Y., McOverton, J., Peterson, A.T., Phillips, S., Wisz, M.S., Zimmermann, N.E., 2006. Novel methods improve prediction of species distributions from occurrence data. Ecography 29, 129 151. ESRI, 2008. ArcMap 9.3. Environmental Systems Research Institute, Inc., USA. Ferrier, S., Pressey, R.L., Barret, T., 2000. A new predictor of the irreplaceability of areas for achieving a conservation goal, its application to real-world planning, and a research agenda for further renement. Biol. Conserv. 93, 303325. Fielding, A.H., Bell, J.F., 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ. Conserv. 24, 3849. Frankham, R., 1997. Do island populations have less genetic variation than mainland populations. Heredity 78, 311327. Fraser, D.J., Bernatchez, L., 2001. Adaptive evolutionary conservation: towards a unied concept for dening conservation units. Mol. Ecol. 10, 27412752. Grivet, D., Sork, V.L., Westfall, R.D., Davis, F.W., 2008. Conserving the evolutionary potential of California valley oak (Quercus lobata Ne): a multivariate genetic approach to conservation planning. Mol. Ecol. 17, 139156. Hazevoet, C.J., 2010. Sixth report on birds from the Cape Verde Islands, including records of 25 taxa new to the archipelago. Zool. Caboverdiana 1, 344. Hglund, J., 2009. Evolutionary Conservation Genetics, rst ed. Oxford University Press, Oxford, New York. IUCN, UNEP-WCMC, 2011. The World Database on Protected Areas (WDPA): January 2011. United Nations Environment Programme World Conservation Monitoring Centre, Cambridge, UK. Jackson, S.F., Kershaw, M., Gaston, K.J., 2004. The performance of procedures for selecting conservation areas: waterbirds in the UK. Biol. Conserv. 118, 261270. Jarvis, A., Reuter, H.I., Nelson, A. Guevara, E., 2006. Hole-lled seamless SRTM data, Version 3. International Centre for Tropical Agriculture (CIAT), Colombia. <http://srtm.csi.cgiar.org/> (accessed September 2008). Kahindo, C., Bowie, R.C., Bates, J.M., 2007. The relevance of data on genetic diversity for the conservation of Afro-montane regions. Biol. Conserv. 134, 262270. Kremen, C., Cameron, A., Moilanen, A., Phillips, S.J., Thomas, C., Beentje, H., Dranseld, J., Fisher, B.L., Glaw, F., Good, T.C., Harper, G.J., Hijmans, R.J., Lees, D.C., Louis, E., Nussbaum, R.A., Raxworthy, C.J., Razampahanana, A., Schatz, G.E., Vences, M., Vieites, D.R., Wright, P.C., Zjhra, M.L., 2008. Aligning conservation priorities across taxa in Madagascar with high-resolution planning tools. Science 320, 222226. Le Corre, V., Kremer, A., 2003. Genetic variability at neutral markers, quantitative trait loci and trait in a subdivided population under selection. Genetics 164, 12051219. Lobban Jr., R.A., Saucier, P.K., 2007. Historical dictionary of the Republic of Cape Verde, fourth ed. In: Woronoff, J. (Ed.), Historical Dictionaries of Africa fourth ed., vol. 104 The Scarecrow Press, Inc., Maryland, USA. Lynch, M., Pfrender, M., Spitze, K., Lehman, N., Hicks, J., Allen, D., Latta, L., Ottene, M., Bogue, F., Colbourne, J., 1999. The quantitative and molecular genetic architecture of a subdivided species. Evolution 53, 100110. Liu, C., Berry, P.M., Dawson, T.P., Pearson, R.G., 2005. Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28, 385393. MAAP-DGA, 2012. Protected Areas Project. Ministrio do Ambiente, Agricultura e Pescas Direco Geral do Ambiente. <http://www.areasprotegidas.cv/ index.php?option=com_docman&Itemid=34>.

286

R. Vasconcelos et al. / Biological Conservation 153 (2012) 276286 Schleich, H.-H., 1987. Herpetofauna Caboverdiana. Spixiana 12, 175. Setiadi, M.I., Hamidy, A., Abidin, Z., Susanto, D., Brown, R.M., Peterson, A.T., Li, X., Evans, B.J., 2009. Genetic structure of herpetofauna on Halmahera Island, Indonesia: implications for Aketajawe-Lolobata National Park. Conserv. Biol. 242, 553562. Smith, T.B., Holder, K., Girman, D., OKeefe, K., Larison, B., Chan, Y., 2000. Comparative avian phylogeography of Cameroon and Equatorial Guinea mountains: implications for conservation. Mol. Ecol. 9, 15051516. Soul, M.E. (Ed.), 1986. Conservation Biology: The Science of Scarcity and Diversity. Sinauer Press, Sunderland, Massachusetts. Tear, T.H., Kareiva, P., Angermeier, P.L., Comer, P., Czech, B., Kautz, R., Landon, L., Mehlman, D., Murphy, K., Ruckelshaus, M., Scott, J.M., Wilhere, J., 2005. How much is enough? The recurrent problem of setting measurable objectives in conservation. Bioscience 55, 835849. Thomassen, H.A., Buermann, W., Mil, B., Graham, C.H., Cameron, S.E., Schneider, C.J., Pollinger, J.P., Saatchi, S., Wayne, R.K., Smith, T.B., 2010. Modeling environmentally associated morphological and genetic variation in a rainforest bird, and its application to conservation prioritization. Evol. Appl. 3, 116. USGS, 2009. Vegetation Indices 16-Day L3 Global 250 m. LP-DAAC. United States Geological Survey. <https://lpdaac.usgs.gov/lpdaac/products/ modis_products_table/vegetation_indices/16_day_l3_global_250m/myd13q1>. Vasconcelos, R., Carranza, S., Harris, D.J., 2010. Insight into an island radiation: the Tarentola geckos of the Cape Verde archipelago. J. Biogeogr. 37, 10471060. Vasconcelos, R., Perera, A., Geniez, P., Harris, D.J., Carranza, S., 2012. An integrative taxonomic revision of the Tarentola geckos (Squamata, Phyllodactylidae) of the Cape Verde islands. Zool. J. Linnean Soc. 164, 328360. Vasconcelos, R., Brito, J.C., Carranza, S., Harris, D.J., in press. Review of the distribution and conservation status of the terrestrial reptiles of the Cape Verde Islands. Oryx. van Teeffelen, A., Moilanen, A., 2008. Where and how to manage: optimal selection of conservation actions for multiple species. Biodiver. Inform. 5, 113. Wei, J., Leberg, P., 2002. A GIS-based approach for assessing the regional conservation status of genetic diversity: an example from the southern Appalachians. Environ. Manage. 29, 531544. Whittaker, R.J., Fernndez-Palacios, J.M., 2007. Island Biogeography Ecology, evolution, and conservation, second ed. Oxford University Press, Oxford, New York. Williams, P.H., Arajo, M.B., 2000. Using probability of persistence to identify important areas for biodiversity conservation. Proc. R. Soc. London B 267, 1959 1966. Wilson, A., Arcese, P., Keller, L.F., Pruett, C.L., Winker, K., Patten, M.A., Chan, Y., 2009. The contribution of island populations to in situ genetic conservation. Conserv. Genet. 10, 419430. Wisz, M.S., Hijmans, R.J., Li, J., Peterson, A.T., Graham, C.H., Guisan, A., NCEAS Predicting Species Distribution Working Group, 2008. Effects of sample size on the performance of species distribution models. Divers. Distrib. 14, 763773. Wright, R.G., Mattson, D.J., 1996. The origin and purpose of national parks and protected areas. In: Wright, R.G. (Ed.), National Parks and Protected Areas. Blackwell Science, Cambridge, MA, pp. 314.

Maiorano, L., Falcucci, A., Garton, E.O., Boitani, L., 2007. Contribution of the Natura 2000 network to biodiversity conservation in Italy. Conserv. Biol. 21, 1433 1444. Margules, C.R., Stein, J.L., 1989. Patterns in the distributions of species and the selection of nature reserves: an example from Eucalyptus forests in southeastern New South Wales. Biol. Conserv. 50, 219238. Marmion, M., Parviainen, M., Luoto, M., Heikkinen, R.K., Thuiller, W., 2009. Evaluation of consensus methods in predictive species distribution modelling. Divers. Distrib. 15, 5969. Miller, R.L., 1993. A call for conservation: national park and protected area development in Cape Verde. Courier Forschungsinstitut Senckenberg 159, 25 32. Miralles, A., Vasconcelos, R., Harris, D.J., Perera, A., Carranza, S., 2010. An integrative taxonomic revision of the Cape Verdean skinks (Squamata, Scincidae). Zool. Scr. 40, 1644. Moilanen, A., Kujala, H., Leathwick, J.R., 2009. The zonation framework and software for conservation prioritization. In: Moilanen, A., Wilson, K.A., Possingham, H.P. (Eds.), Spatial Conservation Prioritization Quantitative Methods and Computational Tools. Oxford University Press, Oxford. Moilanen, A., Kujala, H., 2008. Zonation Spatial Conservation Planning Framework and Software. User Manual. Helsinki: Atte Moilanen/Metapopulation Research Group. <http://www.helsinki./bioscience/consplan/software/Zonation/ index.html>. Moritz, C., 2002. Strategies to protect biological diversity and the evolutionary processes that sustain it. Syst. Biol. 51, 238254. Neel, M.C., Cummings, M.P., 2003. Genetic consequences of ecological reserve design guidelines: an empirical investigation. Conserv. Genet. 4, 427439. OMeally, D., Colgan, D.J., 2005. Genetic ranking for biological conservation using information from multiple species. Biol. Conserv. 122, 395407. Pearman, P.B., DAmen, M., Graham, C.H., Thuiller, W., Zimmermann, N.E., 2010. Within-taxon niche structure: niche conservatism, divergence and predicted effects of climate change. Ecography 33, 9901003. Phillips, S.J., Anderson, R.P., Schapire, R.E., 2006. Maximum entropy modeling of species geographic distributions. Ecol. Model. 190, 231259. Phillips, S.J., Dudk, M., 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31, 161175. Pressey, R.L., 1994. Ad hoc reservations forward or backward steps in developing representative reserve systems. Conserv. Biol. 8, 662668. Quigg, P.W., 1978. Protecting Natural Areas: An Introduction to the Creation of National Parks and Reserves. National Audubon Society, New York. Rissler, L.J., Hijmans, R.J., Graham, C.H., Moritz, C., Wake, D.B., 2006. Phylogeographic lineages and species comparisons in conservation analyses: a case study of California herpetofauna. Am. Nat. 167, 654666. Rodrigues, A.S.L., Gaston, K.J., 2001. How large do reserve networks need to be? Ecol. Lett. 4, 602609. Rodrigues, A.S.L., Akakaya, H.R., Andelman, S.J., Bakarr, M.I., Boitani, L., Brooks, T.M., Chanson, J.S., Fishpool, L.D.C., da Fonseca, G.A.B., Gaston, K.J., Hoffmann, M., Marquet, P.A., Pilgrim, J.D., Pressey, R.l., Schipper, J., Sechrest, W., Stuart, S.N., Underhill, L.G., Waller, R.W., Watts, M.E.J., Yanl, X., 2004. Global gap analysis: priority regions for expanding the global protected-area network. Bioscience 54, 10921100.