The Cerrado is located in central-eastern South America. It is covered by an heterogeneous mosaic of savannic and forest vegetation, including grasslands, shrublands and riverine forests, consisting of a gradient of altitude and vegetation density (Eiten 1972, 1982). Covering over 2.5 million km2, the Cerrado is renowned for its high species richness and endemism that places it as the planet's most diverse savannah. However, during the past 40 years their land has been converted mainly into crops and pastures, leading to an intense process of destruction and fragmentation of the vegetation (Klink and Machado 2005). Currently, the widest remnants of natural vegetation are mainly concentrated in the northern portion (Fig. 1). According to recent estimates, there are only 34 % of the original vegetation left and this is expected to disappear in 30 years if current rates of deforestation are maintained in the region, where traditional cultures are giving place to modern mechanized crops such as soybeans, cotton, corn, sorghum and sunflower (Machado et al. 2004). There is not a consensus about the delimitation of the Cerrado. However, since one of the main objectives of this study is to provide tools for decision-making related to conservation, we chose here to use the biome boundaries that are also adopted by the federal government's policies (IBGE 2004).
Data Used and Pre-processing
Planning Units Planning units (PUs) are subdivisions of the study area into small spatially explicit units. Among many possible ways of obtaining PUs, we used a hydrosheds arrangement built from SRTM (Shuttle Radar Topography Mission; Hydrosheds, hydrosheds.cr.usgs.gov/index.php). This is the same database used by Brazilian government to set priority areas for conservation in Cerrado and Pantanal Biomes (MMA 2012, unpublished data). The use of sub basins as PUs has many advantages over other arrangements such as grids or hexagons: firstly, they have natural and biogeographically meaningful limits; secondly, they allow an hierarchical structure of basins within basins, which is very useful to switch scales and adjust data and results to different needs. To account for the complementarity principle of systematic conservation planning, strictly protected areas (IUCN categories I to IV) were included as PUs, using their actual boundaries regardless of the basin subdivision to design PUs. We only included protected areas wider than 350 km2 to keep PUs sizes compatible with the scale of study and compatible to the official map of priority areas for conservation of the Cerrado, published by the Ministry of Environment. Twenty-three out of 108 protected areas were considered in the gap analysis, covering 50,640 out of 56,223 km2 of IUCN categories I-IV protected areas
Fig. 1 The Cerrado and its relation with other biomes (inset). Distribution of the Cerrado vegetation remnants (gray) and Protected Areas (PAs) greater than 350 km2 (black)
in the Brazilian Cerrado. To define the area available within each PU, we overlaid the official map of extent of natural vegetation in the Cerrado in 2010 with PUs (data available from siscom.ibama.gov.br/monitorabiomas/cerrado/index.htm), and excluded any PU having no remnants of natural vegetation.
Conservation Cost The cost for each PU was obtained from WWF (Soares et al. 2012). The database was built by the Conservation Science Team based on potential future deforestation, using Land Change Modeler module of Idrisi Selva. Distance to roads, to cities, to infrastructure and to previously deforested areas were included as driver to changes in land cover from 2002 to 2010, and then applied to 2010 natural vegetation map to predict which areas are more likely to be deforested in the next 10 years.
Focal Species Eighty-two out of 209 amphibian species known to occur in the Cerrado (Valdujo et al. 2012) were selected as focal species. The criteria were based on endemism, range size (both obtained from Valdujo et al. 2012) and level of tolerance to anthropogenic alterations in habitat quality (two classes: tolerant and nottolerant; species were classified based in our field experience, so that species commonly seen in disturbed areas were considered as tolerant). We used both endemism and extent of distribution as independent criteria because some species are endemic to the Cerrado but have a wide range within this biome, whereas some other species are range restricted (e.g. <60,000 km2) but occur in a transition zone between Cerrado and Atlantic Forest, and so they are not endemic to the Cerrado (Valdujo et al. 2012). Since we were prioritizing among natural areas within the Cerrado, widespread species do not add to the final solution, and neither do species that can tolerate habitat degradation.
Species Distribution Models We prepared geographic distribution maps for all 82 species, using distribution models constructed through the Maximum Entropy algorithm – MAXENT (Elith et al. 2006; Phillips and Dudik 2008). We included as predictors elevation and all 19 bioclimatic variables with a 10 arc-min spatial resolution provided by Worldclim (Hijmans et al. 2005). For each species we used the mean model of 20 runs and converted probabilistic models to binary models using the 10 percentile training presence logistic threshold. Distribution maps were lately validated by a group of experts during a workshop organized by the Ministry of Environment and WWF aiming to identify priority areas for biodiversity conservation in the Cerrado, in 2011, following the procedure recommended by Graham and Hijmans (2006). The distribution map for each species was superimposed onto the PUs' map in order to calculate how much of its distribution area is contained in each PU. All distribution maps were overlaid to obtain the richness surface of endemic species of amphibians in the Cerrado.
Evolutionary History Prioritization In some cases the outcome of area prioritization through SCP analyses fails to meet all targets. To ensure that at least the most important species meet their targets, it is possible to set a penalty factor (SPF) for each species that penalizes solutions more heavily when not achieving these targets. We assigned SPF based on both threat and phylogeny, using ED scores (Evolutionarily Distinctiveness) obtained from Isaac et al. (2012), ranging from 4669 to 17,903.
Mapping Total Evolutionary Distinctiveness We calculate the total ED of each PU by summing the value of all species occurring in it. As ED is highly correlated with richness, here we used a weighted value, obtained by dividing summed ED by richness in each PU.
Table 1 Criteria for the definition of quantitative targets (percentage of range size already under legal protection), according to species range size
Table 2 Gap category, according to the percentage of quantitative target reached
Gap Analysis To evaluate the conservation status of each of the focal species we performed a gap analysis (Rodrigues et al. 2003, 2004). This analysis consists of overlaying species distribution maps and protected areas to calculate how much of the quantitative target set for each species is already under legal protection. Spatial data for Brazilian protected areas were obtained from the Ministry of Environment website (mapas.mma.gov.br/i3geo/mma/openlayers.htm?u3n6kqkh7ajn4igbe 5jilhka56). Targets were set to 20–80 % according to range size (Table 1). Those for which only up to 20 % of its conservation goal has been reached were considered “gap species”. The reaching from 20 to 90 % of the target were considered “partial gaps”, and above 90 % the species was considered “covered” (Rodrigues et al. 2003, 2004) (Table 2).
To select areas and define a conservation scenario for Cerrado amphibians we used the conservation planning software Marxan available online ( uq.edu.au/marxan/index.html; Ball and Possingham 2000). Marxan uses a simulated annealing optimization algorithm for minimizing costs of achieving conservation targets. Planning units defined by protected areas were assigned to status 2, “reserved”. We set to 10,000 runs with 1 million iterations each run, temperature decreases = 10,000, and boundary modifier = 0.2. The identification of priorities for expanding the current network of protected areas was based on measures of “biological significance” (irreplaceability) of each PU within the study area.
Only to assist the identification of some areas within the basins we used geomorphological units denominations (IBGE 2011).