Species richness as criterion for global conservation area placement leads to large losses in coverage of biodiversity

To quantify and compare species coverage in priority areas for conservation identified using species richness as opposed to approaches that use individual species range maps.


| INTRODUCTION
Protected areas are the primary mechanism to protect species from extinction (Margules & Pressey, 2000). Global terrestrial protected area coverage has gradually increased from around 10% in the mid-1990s to 15.4% in 2014(Juffe-Bignoli et al., 2014 with a target of 17% terrestrial area protection by 2020 (Convention on Biological Diversity 2010). Despite the increase in terrestrial land placed under formal protection, biodiversity continues to be lost at an alarmingly rapid rate with estimates suggesting that current extinction rates are approximately 1,000 times higher than background rates (Pimm et al., 2014) and trends in biodiversity indicators continue to decline as they have for at least the past four decades (Butchart et al., 2010; Secretariat of the Convention on Biological Diversity 2014). Additionally, the current protected area network fails to represent all biodiversity that needs protection . The current protected area network, in fact, protects less than 10% of the ranges for 60% of rare amphibians, 50% of rare birds and 44% of rare mammals (Cantú-Salazar, Orme, Rasmussen, Blackburn, & Gaston, 2013) and entirely omits 27% of threatened amphibians, 20% of threatened birds and 14% of threatened mammals (Pimm et al., 2014).
Understanding broad scale biodiversity patterns and identifying priority areas for conservation is an important focus of conservation science (Margules & Pressey, 2000). Understanding where the most important areas for biodiversity are located globally can help allocate limited conservation resources more effectively (Moilanen, Wilson, & Possingham, 2009). Much of the work to understand broad scale biodiversity patterns has used species richness of all, small-range and threatened species to map 'hot spots' of species richness Ceballos & Ehrlich, 2006;Grenyer et al., 2006;Jenkins, Pimm, & Joppa, 2013;Jenkins, Van Houtan, Pimm, & Sexton, 2015;Orme et al., 2005;Pimm et al., 2014). Some of these studies have gone on to suggest that small-range species richness is a suitable surrogate to identify priority areas for conservation (Ceballos & Ehrlich, 2006;Orme et al., 2005), and one has identified global conservation priority areas based on small-range vertebrate species richness (Jenkins et al., 2013). However, prioritization approaches based on richness do not always lead to the highest coverage of species in reserve networks (Albuquerque & Beier, 2015;Justus & Sarkar, 2002;Margules, Pressy, & Williams, 2002;Williams et al. 2006aWilliams et al. , 2006b. Despite this, species richness, particularly richness of small-range species, continues to be used as a measure of conservation value in identifying global biodiversity hotspots (Ceballos & Ehrlich, 2006;Grenyer et al., 2006;Jenkins et al., 2013;Orme et al., 2005;Pimm et al., 2014) and priority areas for conservation (Jenkins et al., 2013;Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000). At the global scale, we currently lack evidence showing the differences in species coverage when priority areas are identified using species richness rather than individual species range maps.
In this study, we quantify and compare the coverage of species ranges from six different spatial conservation prioritization analyses to understand whether alternative analysis structures are suitable for well-informed conservation planning. The analyses considered are based on: (1) species richness, the total number of different species present; (2) weighted range size rarity, a richness measure corrected for individual species' weights and global range size; and (3) complementarity using individual species range maps. We included priority areas based on weighted range size rarity to assess whether it could be a practical alternative to both species richness and a complementarity-based analysis. Prioritizing the landscape based on species richness offers a fast approach with low data and computing requirements. Knowing that species richness patterns are driven by wide-ranging, common species (Brooks et al., 2006;Eken et al., 2004) and that richness-based priority areas fail to maximize species coverage (Albuquerque & Beier, 2015;Justus & Sarkar, 2002;Margules et al., 2002;Williams et al. 2006aWilliams et al. , 2006b, this may still be an attractive option when decisions need to be made quickly with limited information and resources. Basing priority areas on range maps for several thousand individual species in a complementarity-based analysis is extremely data intensive and the necessary computing requirements can easily be a limiting factor. Like species richness, a prioritization based on weighted range size rarity requires modest computing power, but unlike species richness, it should better highlight areas where there is a relatively high density of threatened and/or small-range species. Previous studies have promoted the use of rare and/or threatened species richness as a surrogate for biodiversity (see e.g., Jenkins et al., 2013;Myers et al., 2000;Orme et al. 2005). We implement additional analyses based on species richness subsets including (1) small-range species richness; (2) threatened species richness; and (3) richness of species that are both small-range and threatened. These were conducted to evaluate this claim by assessing species coverage in priority areas based solely on the use of rare and/or threatened species richness. We assessed the difference in coverage achievable for all species, small-range species and threatened species in the 17% highest priority areas, representing a theoretical protection of 17% of terrestrial land area as proposed by the Convention on Biological Diversity in Aichi Target 11 (Convention on Biological Diversity 2010).

| Data
We based our analysis on a set of 21,075 species range maps included in the IUCN Red List of terrestrial vertebrates (IUCN 2014 2014). Hence, they were not included in the analysis. Distribution data for species were available as GIS polygons, covering known or inferred areas where species occur (IUCN 2014). These distribution polygons are in practice positioned somewhere between the extent of occurrence and the true area of occupancy of the species. All range maps were converted to latitude/longitude coordinate system and rasterized to global high resolution grids (latitude/longitude coordinate system, harmonized to resolution 0.01667 degrees equalling 1.7 km at the Equator) using ArCgis 10.1 (ESRI 2012) for use in the spatial prioritization analysis (see below).
The original polygon maps contained five discreet categories indicating the likelihood of the species occurring in that location (Brooks et al., 2016;IUCN 2014). In the rasterizing process, we translated these to four categories with a continuous scale from 1 to 0 and as-

| Species richness and weighted range size rarity
We produced 12 species richness layers in total: all amphibians, birds and mammals; small-range amphibians, birds and mammals; all critically endangered, endangered and vulnerable species; and small-range critically endangered, endangered and vulnerable species. Species richness was calculated from the rasterized IUCN Red List species range maps (IUCN 2014) by summing the values of each raster grid cell in the IUCN range maps corresponding to each richness layer (i.e., to produce a richness layer for small-range amphibians, we summed the IUCN range maps for small-range amphibians).
Prior to calculating species richness, we reclassified the pixel values of each raster (described above) from either 0.5 to 1.0 or 0.1 to 0.0 using the Reclassify function in ArCgis 10.1 (ESRI 2012) to base the species richness calculation on presence/absence data rather than probability of occurrence.
The weighted range size rarity (Williams, Gibbons et al., 1996) map included in this study was created automatically from the Zonation spatial prioritization analysis  based on 21,075 species range maps. Weighted range size rarity for each cell, i, is defined as where w j is the weight assigned to species j in the prioritization and q ij is the fraction of species j's range falling within cell i . This measure lowers the contribution of wide ranging species to overall species richness and highlights the areas that have a relatively high proportion of narrow-range species.

| Spatial conservation prioritization
We used the zonAtion software for spatial conservation prioritization to identify priority areas for conservation action. zonAtion implements analytical methods for broad-scale high-resolution spatial prioritization and works by creating a hierarchical ranking which starts from the full landscape and generates a priority ranking by iteratively ranking and removing the grid cells in the order which cause the least loss in aggregate conservation value (Lethomäki & Moilanen, 2013;Moilanen et al., 2005Moilanen et al., , 2011Moilanen et al., , 2014Pouzols et al., 2014).
During the prioritization process, the landscape-level coverage of each feature (i.e., species) declines by a small fraction after the ranking and subsequent removal of each grid cell in which the feature occurs.
This loss is tracked and is used to balance the relative importance of the remaining occurrences of features (Moilanen et al., 2005(Moilanen et al., , 2011. This balancing operation allows zonAtion to retain all or almost all features throughout the prioritization. There are several options for how zonAtion aggregates marginal loss of conservation value. Here we used the additive benefit function (ABF) variant of analysis (Moilanen, 2007), which bases ranking on minimization of expected extinction rates summed across species (Moilanen et al., 2011. As the ABF has an additive (sum) structure, it tends to emphasize locations where many species (both common and rare) occur (Moilanen et al., 2011.

| Analysis variants
We conducted six separate spatial conservation planning assessments (Table 1). In all assessments, the content of the data was restricted to non-Antarctic terrestrial areas using a binary land/water mask (downloaded from and processed by WorldGrids (http://www.worldgrids. org) and to the areas falling outside of medium intensive and intensive cropland, urban and peri-urban areas for the year 2000 (Van Asselen & Verburg, 2013). The analysis area in the small-range species richness, threatened species richness and small-range and threatened species richness prioritizations were further restricted to the combined extent of the species richness layers included in that assessment using a hierarchical analysis mask (i.e., the analysis area in the prioritization based on small-range species richness was restricted to the combined extent of richness layers for small range amphibians, birds and mammals). In contrast to a binary analysis area mask, a hierarchic mask can contain several different values where grid cells with a low value will be ranked and removed first and grid cells with a higher value will be ranked and removed last (i.e., forced into the top fraction of the priority areas) regardless of relative conservation value . Using the hierarchic mask ensured that the overall analysis area was the same in all analysis variants even though the combined extent of the small-range and/or threatened species richness layers was less than the total terrestrial surface. In the analyses based on small-range and/or threatened species richness, areas covered by the richness layers were given a higher value in the hierarchic mask while other terrestrial areas not covered by the richness layers were given a lower value.
We used a condition transformation (Leathwick, Moilanen, Ferrier, & Julian, 2010;Moilanen et al., 2014)  We conducted our assessment of species coverage at the 17% target for terrestrial land protection set under Aichi target 11 (CBD 2010). The analysis area for priorities based on species richness, species range maps and weighted range size rarity are identical and cover the entire terrestrial surface minus the extent of agriculture and urban areas, and so in these analyses, the top 17% priority areas correspond to 17% of the total analysis area. The analysis area for priorities based on small-range species richness, threatened species richness and small-range and threatened species richness is 7.44%, 86.84% and 87.17%, respectively, of the total analysis area. In the small-range species richness analysis, species coverage was assessed in the full extent of the small-range species richness layers as they covered only a fraction of the area proposed for terrestrial land protection in Aichi target 11. In the threatened species richness and small-range and threatened species richness analyses, we assessed species coverage at the fraction of the extent covered by the richness layers that is equal to 17% of the full global analysis area. For the analysis based on threatened species richness, this equals 19.57% of the extent of the analysis area.
For the analysis based on small-range and threatened species richness, this equals 19.50% of the analysis area.
The original species richness and weighted range size rarity analyses do not contain any species-specific information, and, thus we are not able to determine species coverage in the priority areas based on these datasets. To solve this issue, we conducted a solution loading analysis in zonAtion. With a solution loading analysis, it is possible to evaluate the performance of alternative data in a priority ranking developed under different analysis settings and/or data. In this study, we loaded the output rank maps from the species richness-based and   In all prioritizations, the maximum mean species coverage at the start of the analysis was 80% (Figures 2 and 3). This is a result of the adjustment to the species ranges that occurred after applying the  ( Figure 3, Table 2). The mean species coverage in areas identified using threatened species richness only was between 15% and 38% lower than in areas identified using both small-range and threatened species richness ( Table 2).
The mean species coverage in the 17% highest priority areas based on small-range and threatened species richness is similar to that in the areas based on total species richness, and may thus serve as a suitable surrogate for this measure (Figures 2 and 3, Table 2). However, smallrange and threatened species richness is unlikely to be a suitable surrogate for individual species range maps or weighted range size rarity, as these measures were able to retain an overall higher mean species coverage in the top priority areas (Figures 2 and 3, Table 2).
Differences between methods persist also when considering the number of species left without any protection in the top 17% of the terrestrial landscape. Of the four analyses based on species richness, the proportion of species left without any coverage at the 17% target for terrestrial land protection was lowest in the small-range and threatened species richness analysis (Table 3) (Table 3).

| DISCUSSION
As conservation resources are inadequate, it is vital to use the most appropriate data and tools to maximize species protection. Our results suggest that basing global priority areas for conservation on species richness layers will lead to lower mean coverage of species compared to when complementarity methods are used to identify priority areas.
While the quality of the information on the distribution of species at the global level still needs to be improved (Di Minin & Toivonen, 2015;Meyer, Kreft, Guralnick, & Jetz, 2015), identifying priority areas using species range maps (Brooks et al., 2015) or newly developed species distribution models (Jetz, McPherson, & Guralnick, 2012) in spatial conservation prioritization will maximize coverage of species distributions. Sophisticated methods to identify priority areas for conservation actions based on individual species range or distribution models are already available for this purpose (e.g., Zonation- Moilanen et al., 2005Moilanen et al., , 2014Marxan-Ball, Possingham, & Watts, 2009).
Regional and global scale studies have shown that complementaritybased methods are better able to represent more species in a smaller area (Kati et al., 2004;Reyers, Van Jaarsveld, & Krïger, 2000;Virolainen et al., 2000;Williams, Gibbons et al., 1996). Still, recent literature concerning global conservation priority areas suggests that small-range and/or threatened species richness could be used to locate priority areas for conservation (Ceballos & Ehrlich, 2006;Jenkins et al., 2013).
In fact, recent work by Jenkins et al. (2013) suggests that conservation priority areas based on small-range vertebrate richness are able to also protect large proportions of all terrestrial amphibians, birds and mammals and that focusing conservation efforts on richness of small-range species can help prevent extinctions. In contrast, we show here that species richness, including small-range species richness and threatened species richness, is not an effective surrogate for other biodiversity. Global priority areas identified with species richness layers, in fact, had much lower average species range coverage compared to priority areas identified based on complementarity methods and individual species range maps.
There are, of course, significant trade-offs in data quality and availability that need to be considered when determining which criterion to use to identify conservation priority areas (Di Minin & Toivonen, 2015).
The three approaches (species richness, species range maps, weighted range size rarity) explored in this study each have different data requirements which may influence the decision to use one approach over another. The species richness hotspot method is convenient as species richness is relatively easy to measure (Fleishman, Noss, & Noon, 2006), spatial data are not needed across the full area to calculate species richness locally, and broad-scale analyses can be performed with relatively little computing power or sophistication. However, species richness reduces the highly multidimensional entity of biodiversity into a single dimensional measure and removes all species-specific information (Fleishman et al., 2006). Furthermore, species richness patterns are highly influenced by the distributions of common, wide-ranging species (Brooks et al., 2006;Eken et al., 2004); thus, a conservation priority area network based on species richness might over-represent common, wide-ranging species and under-represent rare, small-range species while overall failing to maximize biodiversity coverage.
Weighted range size rarity, which we also assessed here, could be one relatively quick and easy alternative to using species richness to identify priority areas for conservation. Conceptually, weighted range size rarity is a richness like measure where the contribution of each species has been adjusted to reflect the species weight and the global range (or population) size prior to summing across all individual species (Williams, Prance, Humphries, & Edwards, 1996). Our results show that priority areas based on the weighted range size rarity map achieve average species coverage that is similar to that of priority areas based on individual species range maps. Nevertheless, the analysis based on weighted range size rarity fails to cover all species, because range-size rarity still is a one-dimensional simplification of biodiversity, albeit a more informative one than species richness. Consequently, priority areas based on individual species range maps still perform best.
As with previous global conservation planning assessments, a number of caveats need to be highlighted in this study. Species range maps such as the ones used in this analysis are susceptible to commission errors (when a species is mistakenly thought to be present) and omission errors (when a species is mistakenly thought to be absent), which may have affected our estimates of species coverage (Rondinini, Wilson, Boitani, Grantham, & Possingham, 2006). However, applying a high-quality land-use and land-use change model to improve the individual species range maps should certainly help reduce both commission and omission errors because selected areas are less likely to be those where species are absent owing to anthropogenic habitat loss.
At the same time, availability of broad-scale high-resolution species distribution models is increasing (Jetz et al., 2012), and our present assessment could be improved in the future when this information becomes freely available.
The species richness layers used in this study were derived from the individual species range maps, implying that these richness layers are of relatively high quality. Nevertheless, we find that using species richness layers in spatial conservation prioritization will potentially lead to decreased coverage for many species. This reinforces that individual species range maps, such as the ones available from the IUCN, or other more refined species distribution models are a more preferable basis for spatial conservation prioritization. Finally, we used fine-resolution rasterized versions of the IUCN species range maps as a small cell size best approximates the shapes of the original polygons. We expect that using a coarser resolution would produce similar prioritization results, as has been found previously (see e.g., Pouzols et al., 2014).
This work evaluated the potential effectiveness of priority areas by examining how well they cover species range maps: this is a basic function of conservation areas as species cannot be protected where they do not occur. Nevertheless, coverage is just one of several ways that can be used to judge the effectiveness of a protected area. In order to deem a protected area successful, other factors may need to be considered, including ecosystem function and community structure (Barnes et al., 2014), ecosystem services (Kukkala & Moilanen, 2016), animal populations and habitat condition, and the effect of habitat loss and degradation in the broader landscape outside the protected area (Leverington, Costa, Pavese, Lisle, & Hockings, 2010). How effective a protected area is can also be evaluated with metrics related to the social benefit, such as protection of cultural values, socioeconomic impacts and effectiveness of the management in protected areas (Eklund & Cabeza, 2016;Leverington et al., 2010). Finding the best areas to protect biodiversity is a challenging task, especially when resources are limited and the human population continues to place expanding pressures on the landscape. Planning for new protected areas, as well as evaluating existing areas, should also take these additional factors into consideration.

| CONCLUSIONS
Our results confirm that identifying priority areas for conservation action using species range maps increases average species coverage as compared to priority areas identified using any form of species richness. The results show that priority areas based on species richness are more aggregated and protect fewer species than those identified using individual species range maps. Priority areas based on weighted range size rarity have a mean species coverage that is similar to the analyses based on individual species range maps. Potentially, this makes the weighted range size rarity layer map an important improvement over the species richness methods used in this study.
However, priority areas based on the weighted range size rarity layer still leave some species without protection, showing that prioritization methods based on individual species range maps are able to cover the largest number of species. We do not claim that the priority areas produced from the IUCN range maps in this study should be the focus of future conservation efforts, as factors such as threats and costs should inform priorities as well. We simply show how using species richness to define conservation priority areas can negatively impact species coverage compared to when priority areas are based on single species range maps. Presently available methods and software for processing large numbers of fine-resolution species distribution maps in spatial prioritization greatly complement and improve upon approaches using species richness alone.