|
nif:isString
|
-
This study was carried out in the Caatinga biome (844,453 km2), arid and semi-arid regions extending across eight states of Brazil: Bahia, Sergipe, Alagoas, Pernambuco, Paraíba, Rio Grande do Norte, Ceará, Piauí, and extreme north of Minas Gerais [26] (Figure 1). Xerophytic vegetation type dominated the Caatinga, characterized by spiny deciduous shrubs and trees in association with succulent plants, cacti and bromeliads [27]. In agreement with Andrade-Lima [28], there are twelve Caatinga types distributed in seven physiognomies and six physical units. Annual rainfall may vary from close to zero to as much as ten times the long-term annual average and deviation from the normal rainfall may be higher than 55%. Usually, 20% of the annual rainfall occurs on a single day and 60% in a single month [28] [29]. Most rain falls between September and March. Average annual rainfall is 644 mm, with a 50-year maximum of 1,131 mm and minimum of 250 mm [30]. Mean annual temperature is 27.6°C.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0092950.g001 Location of Caatinga biome in Brazil, protected areas in the Caatinga biome and the presence data used for modeling. The Species Distribution Modeling (hereafter SDM) for jaguar occurrence in Caatinga biome was generated by the maximum entropy algorithm, as implemented in Maxent software 3.3.3e [31] [32]. Maxent is a recently introduced modeling technique, achieving high predictive accuracy and enjoying several additional attractive properties [32]. The idea of Maxent is to estimate a target probability distribution by finding the probability distribution of maximum entropy (i.e., that is most spread out, or closest to uniform), subject to a set of constraints that represent our incomplete information about the target distribution. When Maxent is applied to presence-only species distribution modeling, the pixels of the study area make up the space on which the Maxent probability distribution is defined [31]. Different studies have demonstrated the utility of species distribution modeling to identify areas of high conservation value, as performed by Maxent [12] or ensemble models [11], with Maxent showing, in general, best performance [11] [33] [34] [35] [36] [37]. Models were generated using presence-only data (N = 62) (Table S1; Figure 1) and environmental variables (Table 1) at a spatial resolution of 0.0083 decimal degree (∼1 km2). We selected functionally relevant variables for the species [38], avoiding the autocorrelation. We considered climatic and topographic factors assumed to be important to determine the jaguar distribution, as previously reported [11] [40]. We add two factors that have been reported to be important to determine jaguar presence in the Caatinga biome: distance from water [41] and precipitation of driest month as reported by local people. All presence records were obtained from National Predator Center (CENAP-ICMBio) database and literature [42] [43]. All runs were set with a convergence threshold of 1.0E–5 with 500 iteractions and with 10,000 background points, auto features, and analysis of variable importance measured by Jackknife, response curves and random seed.
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0092950.t001 Environmental variables used for Species Distribution Modeling (SDM) for jaguar at Caatinga biome, Brazil. *mean of monthly (max temp - min temp).
The SDM was generated by bootstrapping methods with 10 random partitions with replacements using 70% of the dataset for training and 30% for testing models [44]. The average model was cut off by the 10 percentile training presence logistic threshold (0.2613) as it provided the best accurate model for the species occurrence in the biome. We tested the SDM's predictive ability for jaguar occurrence in the Caatinga biome by plotting a new independent dataset not used for modeling (N = 38; Table S2) from recent species occurrence points. The SDM was evaluated by AUC value, binomial probability and omission error [44] [45].
High Priority Areas for Conservation: We used a different approach from that proposed by Sanderson et al. [24] to identify jaguar conservation units. From the SDM, we selected areas equal to or higher than the median suitability value of 0.595, which represents areas of high suitability for jaguar occurrence [11]. Then, we used the percent volume contour (i.e., raster layer representing a probability density distribution) from Kernel tools in Hawth's analysis tools for ArcGis [46] to delimit these areas, which we named as Priority Jaguar Conservation Units (PJCUs) (i.e., continuous areas of high suitability for jaguar occurrence).
Connectivity modeling was performed among PJCUs as proposed by Rabinowitz and Zeller [19]. We defined five predictors (Table 2) for creating the cost surface or permeability matrix (Table 3) and attributed cost values (ranging from 0 – no cost for jaguar movement – to 10 – high cost for jaguar movement) for each according to Rabinowitz and Zeller [19]. Cost values for elevation, the variable that contributed substantially to the SDM, were attributed based on the marginal response curve provided by the SDM (Figure 2). Following the procedures proposed by Rabinowitz and Zeller [19], we used the Cost-Distance function (Spatial Analyst, ArcGis 9.3) to delineate movement cost grids for each PJCU. After, we used the cost-distance grids as inputs for the Corridor function in Spatial Analyst for all proximate pairs of PJCUs, resulting in least-cost corridors among each pair. Then, we used the minimum mosaic method, combining all overlapping corridors to generate the final least-cost corridor model. Finally, differently from Rabinowitz and Zeller [19], we used the cost path function with cost-distance grids and PJCUs as inputs to calculate the least-cost path from a source to a destination. Crossing the least-cost paths to least-cost corridor model we then selected the best routes, hereafter named corridors, for jaguar dispersal through surfaces with no or low cost for movement. In addition, we identified “buffer zones” around PJCUs and corridors.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0092950.g002 Marginal response curve of altitude, the variable that contributed most to the SDM of jaguar occurrence at the Caatinga biome. Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0092950.t002 Geographical databases used for connectivity modeling. Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0092950.t003 Classes of landscape layers and cost values for jaguar movement. Costs values ranged from 0 (no cost for jaguar movement) to 10 (high cost for jaguar movement). BA means barrier for jaguar movement.
For categorizing PJCUs we considered the follow aspects, in order of importance: 1) PJCU size; 2) connectivity, and; 3) jaguar population status [24]. For PJCU size we estimate the smallest continuous area necessary to preserve a viable population of 50 individuals [24] as suggested by Rodriguez-Soto et al. [11]. In brief, we assumed (1) a sex ratio of at least one male every two females [47] [48] and thus counting on 15 males and 35 females, (2) an average home range of 130 km2 for males and 41 km2 for females [41] and (3) a complete overlap of the home range of one male with two females [49]. In this way the smallest continuous area necessary to preserve a viable jaguar population corresponds roughly to 1,700 km2 of high suitability habitats. In this way, PJCUs≥1,700 km2 received three points. Areas smaller than 1,700 km2 but with adequate habitat where jaguar populations can increase if threats were alleviated received two points. Finally, areas that cannot hold a jaguar population but still can function as stepping stone areas received one point. For connectivity, each PJCU received one point for each possible connection. Considering the jaguar population status, we combined the PJCU size previously calculated, with density estimate (1.57±0.43) previously reported by Sollmann et al. [21] (Table 4). Despite other available densities, Sollmann et al. [21] presented a spatially explicit capture-recapture model resulting in more precise estimates [50] than previously published non-spatial estimates [51] [52]. PJCUs containing at least 50 individuals, considering it to be genetically stable for 100 years [24], received three points, PJCUs containing fewer than 50 individuals but still can increase if threats can be reduced [24] received two points. PJCUs where the smaller estimated population is less than 1.0 but still can function as stepping stone areas received one point. Arbitrarily, we defined PJCUs with 8–9 points as high priority, PJCUs between 5–7 points as medium priority and PJCUs with 3–4 points as low priority.
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0092950.t004 Priority Jaguar Conservation Units (PJCUs) identified in the Caatinga Biome. Total area, estimated population size and connectivity were used to prioritize the PJCUs.NA = smaller estimated population is less than 1.0.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}