Man has shaped landscapes for centuries (Vos & Meekes, Reference Vos and Meekes1999). In the last 4 decades socio-economic and lifestyle changes have driven a rural exodus and the abandonment of land throughout many of Europe's rural landscapes (MacDonald et al., Reference MacDonald, Crabtree, Wiesinger, Dax, Stamou and Fleury2000; Höchtl et al., Reference Höchtl, Lehringer and Konold2005). In some cases sociocultural and economic problems have created new opportunities for conservation (Theil et al., Reference Theil, Elkin, Underhill, Pape and Rogacin2005). Land abandonment and degradation, rural depopulation and the perception of the presence of small pockets of intact nature has increased the interest in wilderness throughout Europe, and rewilding has emerged as a novel concept within restoration and conservation management (for a review, see Jepson, Reference Jepson2015). Rewilding aims to restore ecosystem functioning with minimal human intervention (Sandom et al., Reference Sandom, Bull, Canney, Macdonald, Somers and Hayward2012), which cannot be achieved without the reintroduction of species that have key roles in shaping the landscape (McKnight, Reference McKnight2014). In this context a fundamental part of the minimal human intervention is the reintroduction of keystone species (Schadt et al., Reference Schadt, Revilla, Wiegand, Knauer, Kaczensky and Breitenmoser2002; Manning et al., Reference Manning, Gordon and Ripple2009; Sandom et al., Reference Sandom, Bull, Canney, Macdonald, Somers and Hayward2012; Cianfrani et al., Reference Cianfrani, Maiorano, Loy, Kranz, Lehmann, Maggini and Guisan2013). Since 2011 the Rewilding Europe project has been active in turning land abandonment problems into opportunities (Helmer et al., Reference Helmer, Saavedra, Sylvén, Schepers, Pereira and Navarro2015). One of its most emblematic projects aims to reintroduce viable and self-sustaining populations of the European bison Bison bonasus in connected areas across Europe (van de Vlasakker, Reference van de Vlasakker2014). In the Iberian Peninsula (Portugal and Spain) an important goal of Rewilding Europe is to develop natural grazing systems to facilitate the return of iconic and ecologically important species such as the Iberian ibex Capra pyrenaica, the wild rabbit Oryctolagus cuniculus and the Iberian lynx Lynx pardinus, which are considered to be key species in this region.
Modelling the environmental favourableness and landscape connectivity of a territory in the context of a reintroduction programme is particularly challenging because ecological data on the species to be reintroduced are usually not available for the target area. To date, little has been done on the application of species distribution models in planning the recovery of reintroduced species (Chauvenet et al., Reference Chauvenet, Parlato, Gedir, Armstrong, Armstrong, Hayward, Moro and Seddon2015) within the Rewilding Europe framework. However, the first step in any reintroduction programme is to identify favourable areas for a species and the connectivity between those areas. This is required to optimize the success of reintroduction and the establishment and dispersal of the introduced population (Armstrong & Seddon, Reference Armstrong and Seddon2008).
The Iberian ibex is an emblematic species that was once widely distributed in the Iberian Peninsula and French Pyrenees (Pérez et al., Reference Pérez, Granados, Soriguer, Fandos, Márquez and Crampe2002). In the 20th century hunting pressure, habitat loss and outbreaks of sarcoptic mange led to a local reduction, and consequently fragmentation, of wild goat populations (Pérez et al., Reference Pérez, Granados, Soriguer, Fandos, Márquez and Crampe2002). Of the four subspecies described by Cabrera (Reference Cabrera1911) only two are extant: C. pyrenaica hispanica and C. pyrenaica victoriae. Capra pyrenaica lusitanica inhabited the north-western corner of the Iberian Peninsula and went extinct in the last decades of the 19th century, and C. pyrenaica pyrenaica inhabited the Pyrenees and has been considered to be extinct since 2002. However, several translocations were carried out in the last decades of the 20th century and C. pyrenaica currently has a wide distribution in Spain (Acevedo & Cassinello, Reference Acevedo and Cassinello2009). In 1997 individuals of the subspecies C. pyrenaica victoriae were reintroduced in north-west Spain, in Baixa Limia-Serra do Xurés National Park (a transboundary park with Portugal), from the Gredos mountains of central Spain (Moço et al., Reference Moço, Guerreiro, Ferreira, Rebelo, Loureiro, Petrucci-Fonseca and Pérez2006). This subspecies is categorized as Critically Endangered in the Portuguese Red Data Book (Cabral et al., Reference Cabral, Almeida, Almeida, Dellinger, Ferrand de Almeida and Oliveira2005), Near Threatened in the Spanish Red Data Book (Acevedo & Cassinello, Reference Acevedo and Cassinello2009) and Least Concern on the IUCN Red List (Herrero & Pérez, Reference Herrero and Pérez2008). The Iberian ibex is considered to be a mixed feeder (browser and grazer), dependent on plant availability (Moço et al., Reference Moço, Serrano, Guerreiro, Ferreira, Petrucci-Fonseca and Maia2013; Perea et al., Reference Perea, Perea-García-Calvo, Díaz-Ambrona and San Miguel2015). It is also an ecosystem engineer as it creates and modifies habitats, with significant effects on ecosystem functioning (Byers et al., Reference Byers, Cuddington, Jones, Talley, Hastings and Lambrinos2006).
The Western Iberia region (a montado–sierra landscape in Portugal, bordering Spain) is one of five pilot areas selected for the Rewilding Europe project. This is a rural area in the interior of Portugal, where land abandonment and rural depopulation have resulted in the density of people and roads being generally low.
We applied a two-fold analytical framework to address the following questions related to planning the reintroduction of the Iberian ibex within the Rewilding Europe perspective: (1) Are there areas in central Western Iberia that are environmentally favourable for the Iberian ibex? (2) If so, are these areas well connected with each other? (3) Which of these areas favour the establishment and expansion of a viable population of Iberian ibex in a reintroduction programme? Firstly, we modelled the distribution of the Iberian ibex in the Iberian Peninsula using a UTM (Universal Transverse Mercator) grid of 10 × 10 km cells. Secondly, we downscaled the model to central Western Iberia, our area of interest, using a UTM grid of 1 × 1 km cells. Finally, we evaluated the connectivity between the favourable localities in Western Iberia. By quantifying the importance of each individual habitat area it is possible to focus conservation and management actions directly on those areas that have greater importance for overall connectivity.
The Iberian Peninsula is a heterogeneous territory. There are four main climatic types: Atlantic (mild year-round, with abundant rainfall spread throughout the year; typical of the north-west areas), Continental (wide annual temperature range among seasons), Mediterranean (warm and dry summers and mild winters), and Semi-arid (drought and hot summers). These characteristics play a key role in the distribution of biodiversity (Tenorio et al., Reference Tenorio, Juaristi and Ollero2005). The study was focused on the Western Iberia region, in an area of 1,064 km2 in the valley of the River Côa (Fig. 1). At a coarse scale the climate is predominantly dry and Mediterranean, with some Atlantic influence, mainly in the northern and western areas. The temperature is frequently > 40°C in late spring and summer, daily thermal amplitudes may reach 10–15°C and annual precipitation is often < 400 mm. The vegetation is varied and characterized by Quercus suber, Quercus pyrenaica, Quercus rotundifolia and mixed patches interspersed with understorey species such as Erica australis, Pterospartum tridentatum, Halimium alyssoides and Cistus ladanifer. The area is crossed by a diversity of rivers and small streams. Scattered cultivated areas are mainly planted with vines and olive and almond trees. Sympatric ungulates include the semi-wild garrano Equus ferus caballus, the wild boar Sus scrofa and the roe deer Capreolus capreolus. The area has a low human population density and was one of the areas historically occupied by the Iberian ibex, probably C. pyrenaica lusitanica (Cabrera, Reference Cabrera1911).
The variables used to model the distribution of the Iberian ibex were selected based on information about the species’ ecology (Granados et al., Reference Granados, Pérez, Márquez, Serrano, Soriguer and Fandos2001; Pérez et al., Reference Pérez, Granados, Soriguer, Fandos, Márquez and Crampe2002; Acevedo et al., Reference Acevedo, Cassinello, Hortal and Gortazar2007a; Acevedo & Real, Reference Acevedo and Real2012). These variables included gradients related to habitat structure and vegetation productivity, topographical features and human disturbance (Table 1). The habitat structure comprised three broad land use/land cover categories: forest stands, agricultural fields and scrubland. Forest stands are important as they provide cover for the species. Scrubland plays an important role in providing food but can also provide protection if the shrub layer is high enough to cover a standing ibex. The normalized difference vegetation index, recently linked to the body condition of Iberian ibexes (Carvalho et al., Reference Carvalho, Granados, López-Olvera, Cano-Manuel, Pérez and Fandos2015), was used as a proxy for vegetation productivity (Pettorelli, Reference Pettorelli2013). The slope was used as an index of topographic heterogeneity. We also included two variables that accounted for human disturbance: population density and road density. All variables were handled and processed in ArcGIS v. 10.0 (ESRI, Redlands, USA).
Our approach consisted of two main steps. Firstly, we identified favourable areas in the Iberian Peninsula, using logistic regression modelling and a forward–backward stepwise procedure with UTM 10 × 10 km cells. Secondly, we downscaled the favourableness values to Western Iberia, with UTM 1 × 1 km cells. Finally, we evaluated the connectivity between the favourable localities in Western Iberia, using a graph-based method (Fig. 2).
The extent of the study area affects the outputs of species distribution modelling (Barve et al., Reference Barve, Barve, Jiménez-Valverde, Lira-Noriega, Maher and Peterson2011; Acevedo et al., Reference Acevedo, Jiménez-Valverde, Lobo and Real2012). By modelling the species distribution with the third-degree polynomial of the spatial coordinates as predictors (trend surface analysis; Acevedo et al., Reference Acevedo, Jiménez-Valverde, Lobo and Real2012) we delimited the territory adequate to model the distribution of the Iberian ibex. This procedure selected 5,702 UTM 10 × 10 km cells, of which the Iberian ibex was reported in 645 cells. The ecogeographical drivers of distribution of the Iberian ibex were determined on the basis of the locations in which the species currently occurs. Prior to modelling, the database was divided into training (80%, 4,561 grid cells) and validation (20%, 1,141 grid cells) datasets. To avoid problems with multicollinearity we quantified the variance inflation factor in the training dataset and excluded predictors with a variance inflation factor > 10 from the analyses (Montgomery & Peck, Reference Montgomery and Peck1992). Variance inflation factors were calculated for each predictor as the inverse of the coefficient of non-determination of the regression of each predictor against all others, using HH (Heiberger, Reference Heiberger2012) in R v.3.2.3 (R Development Core Team, 2016). The predictors selected after controlling the variance inflation factor were considered in a multiple logistic regression (Hosmer & Lemeshow, Reference Hosmer and Lemeshow1989) in which the presence or absence of C. pyrenaica was used as the response variable. The final model was obtained using a forwards–backwards stepwise procedure based on the Akaike information criterion. Predictions from the logistic regression were included in the favourableness function to make the predicted values independent of the species’ prevalence in the study area (Real et al., Reference Real, Barbosa and Vargas2006). The favourableness values are interpretable in absolute terms and can therefore be used as a threshold to identify favourable and unfavourable areas for the species (Acevedo & Real, Reference Acevedo and Real2012). Predictive performance of the final model was assessed on the validation dataset and includes the discrimination capability (area under the receiver operating characteristic curve, AUC). The resolution at which distribution data are available does not always match with the interests of conservation and management. Changes in the spatial resolution of model predictions (i.e. downscaling: the projection of models built at one resolution to a finer spatial resolution) hold potential benefit for ecology and conservation (e.g. Barbosa et al., Reference Barbosa, Real and Vargas2010). Thus, the final model was downscaled to UTM 1 × 1 km cells in Western Iberia, to carry out the connectivity analyses.
Graph theory has been suggested as an effective and flexible procedure to assess landscape connectivity (Pascual-Hortal & Saura, Reference Pascual-Hortal and Saura2006; Saura & Torné, Reference Saura and Torné2009; Saura & Rubio, Reference Saura and Rubio2010). Within this framework the landscape is described as a set of habitat patches (nodes) and their connecting elements (links; for more details see Saura & Rubio, Reference Saura and Rubio2010). A link represents the functional connectivity (i.e. the likelihood of dispersal between two habitat patches) and can correspond to an ecological/physical corridor or the dispersal ability of the species. All habitat patches with a certain area are considered to be landscape elements. The habitat patch is described by an attribute value, such as habitat area or habitat suitability (Pascual-Hortal & Saura, Reference Pascual-Hortal and Saura2006). In this case the attribute value of the habitat patch was the area quality, which corresponds to the habitat area divided by the favourableness value derived from modelling. To assess functional connectivity specifically in relation to the Iberian ibex it is essential to include the species’ dispersal ability. According to previous studies the mean dispersal distance of the Iberian ibex is 1.25 km (0.7 km, Alados & Escós, Reference Alados and Escós1985; 1.8 km, Escós, Reference Escós1988). The landscape connectivity was evaluated via the probability of connectivity index (PC; Saura & Pascual-Hortal, Reference Saura and Pascual-Hortal2007; Saura & Torné, Reference Saura and Torné2009). This is expressed as the probability that two points arbitrarily located within the landscape are in habitat areas that are accessible from each other (i.e. interconnected; Saura & Pascual-Hortal, Reference Saura and Pascual-Hortal2007). Node importance (d) expresses the importance of each habitat patch for the overall connectivity (Saura & Torné, Reference Saura and Torné2009). The dPC value can then be divided into three components considering the various ways in which a certain landscape element i (patch or link) can contribute to overall habitat connectivity in the landscape:
where dPCintra i is the contribution of patch i in terms of intrapatch connectivity, dPCflux i corresponds to the dispersal flux through the connections of patch i to or from all of the other patches in the landscape when i is either the initial or ending patch of that connection (dPCflux depends both on the attribute value of patch i—all things being equal, a patch with a higher attribute value produces more flux—and on its position within the landscape), and dPCconnector i is the contribution of patch or link i to the connectivity between other habitat patches, as a connecting element or stepping stone between them. dPCconnector is dependent only on the topological location of a patch or link within the landscape.
The connectivity analyses were done only after downscaling the model to the Western Iberia region. To evaluate the importance of habitat patches for landscape connectivity, and thus to identify optimal sites for future Iberian ibex reintroductions, we used Conefor 2.6 and its geographical information system extensions (Saura & Torné, Reference Saura and Torné2009).
Our model facilitated the identification of a set of favourable areas for Iberian ibex in the Iberian Peninsula (Fig. 3). Circa 32% of the Peninsula has environmental favourableness > 0.5 and the species is already present in 27% of this area. In Western Iberia the environmental favourableness for the species varied throughout the region. The north-centre of the study area was associated with patches of dense shrub layer, rough and sloping ridge tops and valleys, low human population and low road density (Table 2). The final model showed good predictive discrimination at the original 10 × 10 km resolution (AUC = 0.79).
The relative importance of each landscape element for overall habitat connectivity for the Iberian ibex (as quantified by dPC and its three components: dPCintra, dPCflux and dPCconnector) was heterogeneously distributed in the landscape; however, the most important landscape elements were located in the north and central areas of the study area (Table 3; Fig. 4). Our results indicate that landscape element 1 and landscape element 2 (Fig. 4) were the most important landscape elements in maintaining the functionality of habitat connectivity for the Iberian ibex. Landscape element 1 had the highest contribution to the overall connectivity (dPC = 19%) and the highest values of the dPCflux component (12%), followed by landscape element 2 (dPC = 18%), with the highest values of the dPCconnector component (13%; Table 3). Landscape element 1 was well connected to the other habitat patches within the dispersal distance of the Iberian ibex (high dPCflux), whereas landscape element 2 was strategically located to sustain the connectivity between other habitat patches (high dPCconnector). Both landscape elements 3 and 4 had the second highest values of dPCflux and dPCconnector (landscape element 3: dPCflux = 9%, dPCconnector = 8%; landscape element 4: dPCflux = 8%, dPCconnector = 6%). Landscape elements 5 and 6 had the highest values of dPCintra (2% for each; Table 3). Landscape element 2 had the highest dPCconnector, followed by landscape element 3 (Table 3).
We identified favourable areas that could serve as potential nuclei of reintroduction for the Iberian ibex in the Iberian Peninsula, which in turn could lead to establishment of the species and expansion of its distribution. The Western Iberia region is strategically located to enhance the natural expansion of existing populations, including the nearby Spanish population in the Sierra de las Batuecas. Furthermore, the region is characterized by well-connected habitat patches and relatively low human pressure, offering favourable conditions for reintroduction and establishment of the Iberian ibex. The species would have access to continuous habitat to fulfil its ecological requirements, increasing the likelihood of its long-term survival. Previous studies have highlighted the importance of protecting and enhancing existing favourable areas to facilitate the dispersal and survival of the species (Schadt et al., Reference Schadt, Revilla, Wiegand, Knauer, Kaczensky and Breitenmoser2002; Bleyhl et al., Reference Bleyhl, Sipko, Trepet, Bragina, Leitão, Radeloff and Kuemmerle2015). The favourable areas that were identified in the north and centre of Western Iberia could potentially serve as reintroduction nuclei (Fig. 4), and their strategic location would facilitate the movement of populations among the nuclei. The areas with highest favourableness include dense vegetation, well-developed tree cover interspersed with areas with a continuum of dense and sparse shrub layer, and a rich herbaceous layer, which provide refuge and feeding places for species establishment and reproduction. Additionally, forest stands and tall shrub may serve as ecological corridors for range expansion. The species’ preference for low-disturbance regions has already been shown (Acevedo & Cassinello, Reference Acevedo and Cassinello2009). In an area with low disturbance, such as the study area, the main threats to the Iberian ibex are the increasing presence of domestic livestock, which can act as reservoirs of wildlife diseases, and interspecific competition for resources (Acevedo et al., Reference Acevedo, Cassinello and Gortazar2007b).
After assessing the existence of favourable habitats we evaluated whether they were sufficiently well connected to facilitate movement between populations and ensure that suitable patches could be colonized. By ranking the contribution of each landscape element to the maintenance of the overall landscape connectivity we identified high-priority habitat patches for future conservation and management actions. Specifically, our results show that the central area of Western Iberia plays a fundamental role in the dispersal of the Iberian ibex. The landscape elements are strategically located (e.g. in a central position) and are large and well connected to the other habitat patches within the dispersal distance of the species (high intra-patch connectivity). Moreover, they enhance the connectivity between the north, south and eastern parts of the study area, which will promote the natural expansion of the existing populations on the Spanish side. Specifically, both landscape elements 3 and 4 had high contributions to overall connectivity because they are well connected to other habitat patches within the dispersal distance of the Iberian ibex (high dPCflux) and are large enough and centrally located, supporting connectivity between other habitat patches (high dPCconnector). Landscape elements 5 and 6 have the largest total area (AL) and therefore high dPCintra. Landscape elements 2 and 3, in the central part of the study area, have the highest dPCconnector (i.e. they may function as stepping stones between other isolated patches), highlighting the high probability of colonization of these habitat patches. Stepping stones are known to be fundamental for the spread of species (Carranza et al., Reference Carranza, D'Alessandro, Saura and Loy2012; Saura et al., Reference Saura, Bodin and Fortin2014). Fragmentation of the vital core area would probably impede establishment of the Iberian ibex. Connectivity between habitat patches has been shown to be important in securing genetic flux between populations, increasing the efficiency of reintroduction programmes (Olech & Perzanowski, Reference Olech and Perzanowski2002). According to the dispersal characteristics of the Iberian ibex (Escós, Reference Escós1988; Alados & Escós, Reference Alados and Escós1985), natural dispersal to the south and east of the study area is expected, as there are potentially effective connections between the reintroduction sites and the southern habitat patches. By quantifying the importance of each habitat area it is possible to focus conservation and management actions directly in areas that have greater importance for overall connectivity.
We suggest that the approach used here for the Iberian ibex in Western Iberia could be extended to other species and conservation purposes, and is particularly important in reintroduction projects where species data are not available. However, the success of reintroduction programmes depends not only on evaluating favourable and connected areas but also on the ability to manage them appropriately (Jepson, Reference Jepson2015). A main goal of the EU Biodiversity Action Plan is to restore 15% of the EU's degraded or abandoned land, and the creation of a functional ecosystem through rewilding would contribute to achieving this goal. Our proposed methodology for identifying high-priority habitats in target areas is a valuable tool for conservation planning, helping to determine high-priority habitats within target areas. This is particularly significant as rewilding projects are assuming a more widespread role in restoring degraded landscapes (Jepson, Reference Jepson2015).
We are grateful for the support of the Associação de Transumância e Natureza. JC was supported by a PhD grant (SFRH/BD/98387/2013), ES by a postdoctoral programme (SFRH/BPD/96637/2013) of the Fundação para a Ciência e a Tecnologia, (FCT), Portugal, and PA by the Spanish Ministerio de Economía y Competitividad and Universidad de Castilla-La Mancha through a Ramón y Cajal contract (RYC-2012-11970). We thank the University of Aveiro (Department of Biology) and Fundação para a Ciência e a Tecnologia/Ministério da Ciência, Tecnologia e Ensino Superior (FCT/MEC) for the financial support to Centro de Estudos do Ambiente e do Mar Research Unit (CESAM RU) (UID/AMB/50017) through national funds, co-financed by the Fundo Europeu de Desenvolvimento Regional (FEDER) within the PT2020 Partnership Agreement.
RTT and JC conceived and designed the research, analysed data and wrote the manuscript. PA analysed data. All authors contributed to the development of ideas.
Rita Tinoco Torres’s research is focused primarily on the ecology and management of ungulates, and their role as reservoirs of infectious diseases at the human–livestock–wildlife interface. João Carvalho is interested in wildlife management, ecological and evolutionary consequences of selective harvesting in ungulates, multiparasitism, cartography and geographical information systems. Emmanuel Serrano is interested in the impact of infectious diseases and environmental variation on life history traits of individuals and populations, with a particular focus on ungulates. Wouter Helmer has more than 30 years of experience in large-scale rewilding programmes. As member of the Rewilding Europe team his main focus is on the restoration of natural processes and key wildlife species. Pelayo Acevedo is focused on the multidisciplinary framework in which wildlife species and their parasites are studied at local and large spatial scales. Carlos Fonseca has broad interests in wildlife ecology, conservation and management, with a particular interest in ungulates.