Material and Methods

We used the most detailed primary-data species-level phylogenies available for the three major mammalian groups containing aquatic species Cetacea (May-Collado and Agnarsson 2006; May-Collado et al. 2007), Carnivora (Agnarsson et al. 2010), and Afrotheria (Kuntner et al. 2009). The conservation status for 127 aquatic mammals was obtained from the IUCN Red List of Threatened Species database (2010.4–2013.2) and transformed to probability estimates of extinction risk using two of the methods discussed in Mooers et al. (2008) “pessimistic” and “IUCN50”. The “pessimistic” method is an arbitrary transformation that designates a sizable probability of extinction to every category. So, even for the 'least concern' species has a probability of 0.2, which is much higher than in the IUCN 50 scenario (see Table 1) (Mooers et al. 2008). The “IUCN50” is a projection of extinction risk over the next 50 years given current conservation status, proposed by the IUCN. This scenario assumes that species in the 'least concern' category are essentially 'safe', assigning to them low probability of extinction (Mooers et al. 2008) (Table 1). We selected these two transformation methods because they offer contrasting scenarios based on how they treat species that are currently thought to be at relatively low risk. Using these transformation methods we calculated conservation priority measures using the TUATARA module version 1.01 (Maddison and Mooers 2007) in the evolutionary analysis package MESQUITE version 2.75 (Maddison and Maddison 2011). We used the conservation priority methods EDGE (Evolutionary Distinct, Globally Endangered) (edgeofexistence.org), which measures evolutionary distinctiveness (ED) weighted by current IUCN levels of extinction risk. EDGE scores are equivalent to a logarithmic transformation of the product of

Table 1 Extinction probabilities for IUCN levels of imperilment transformed into extinction probabilities using “pessimistic” and IUCN 50 transformations,

as proposed by Mooers et al. 2008

aMooers et al. 2008

bMay-Collado and Agnarsson 2011

a species' evolutionary distinctiveness and the probability it will go extinct (see Isaac et al. 2007). We also use the conservation priority method HEDGE (Heightened EDGE), which like EDGE measure evolutionary distinctiveness by IUCN levels of extinction but additionally considers how future extinction of species will affect the evolutionary distinctiveness of remaining species. In sum, HEDGE estimates the expected terminal branch length of the focal species in light of future extinction risk (Steel et al. 2007). Both conservation methods generate fixed probabilities of extinction as described in Table 1. For more information on how IUCN levels of imperilments are transformed into probabilities of extinction see Moores et al. (2008). Because there are not estimated probabilities of extinction available for data deficient species we arbitrarily assigned an extinction risk score in between the two lowest IUCN extinction categories: least concern and near threatened (Table 1). All four metrics were calculated using both 'raw' branch lengths (estimated by MrBayes) as they contain information on the unique evolutionary information of terminal taxa and ultrametrized trees. Furthermore, with the purpose of comparing this approach to identify conservation priority areas with other commonly used conservation prioritization criteria we calculated evolutionary distinctiveness (ED) (Isaac et al. 2007) also implemented in the TUATARA module and gathered information on species richness.

To identify conservation priority areas (CPAs) we used distribution range maps from the IUCN spatial database (2013) as a baseline to produce species richness, ED, EDGE and HEDGE maps, under both IUCN extinction probabilities transformation methods, pessimistic and IUCN50. The IUCN spatial database depict species' range distribution as polygons based on the extent of occurrence, which is defined as the area contained within a minimum convex hull around species' observations or records. This convex hull or polygon is further improved by including areas known to be suitable or by removing unsuitable or unoccupied areas based on expert knowledge.

For each species the distribution range was converted to a grid system with cells of 5′ × 5′ (approximately 10 × 10 km at the Equator line). This spatial resolution was selected for its practical compromise between intensive computing and a reasonable representation of geographic patterns. Traditionally, a one-degree cell (100 × 100 km) has been used in macroecological analyses, but there is no ecological reason behind the resolution. More importantly, it has been shown that higher spatial resolutions distort the geographical patterns of species richness (Rahbek 2005; Graham and Hijmans 2006). In contrast, lower spatial resolutions minimize the overestimation of distribution ranges, in particular of those species with small range distributions. For example, Rondinini et al. (2011) used a resolution of 300 × 300 m for their estimations of global mammal' species richness.

Specie's presence in each 5′ × 5′ grid cell was assigned with a value of one. The same procedure was repeated to assign the estimated values of ED, EDGE and

HEDGE to the grid cells of each species' occurrence, and maps were calculated by overlying individual grids. For example, the species richness map represents the sum of all presence grids. Under this spatial framework CPAs represent areas with the highest scores due to a high number of species regardless of ED, EDGE, and HEDGE scores or to a few species with high probability scores.

To understand how well these patterns of aquatic mammal conservation priorities are already included in existing MPAs, we used information from the World Database on Protected Areas website (protectedplanet.net/). We calculated the percentage of each species range within all designated MPAs of the world and to preserve areal relationships we first re-projected both the species ranges and the MPAs, to an equal area using Behrmann projection. We then iteratively intersected each marine species range with all the MPAs using the function joinPolys with the operator 'INT' (intersection) from the package PBS-Mapping (version 2.67). Species range percentage within all MPAs was calculated by contrasting the sum of their intersected areas with the total species range area. All analyses were performed in R (R Core Team 2013) and the final maps were created using ArcGIS v10.1.