Correspondence author: Mar Cabeza, Metapopulation Research Group, Department of Biological and Environmental Sciences, PO Box 65 (Viikinkaari 1), FI-00014 University of Helsinki, Finland, Tel.: +358 9191 57743, Fax: +358 9191 57694. E-mail: email@example.com
1Sergio et al. (2006) argue that top predators are justified conservation surrogates based on a case study where raptor presence is associated with high species richness of birds, butterflies and trees.
2We question the methodology as well as the applicability of their results, and clarify differences between surrogates for biodiversity hotspots and surrogates for complementarity. We show that the results from Sergio et al. related to richness hotspots are not fully reliable and that the ability of top predators to identify complementary areas is not demonstrated. Given that complementarity-based surrogate studies have produced mixed results for a variety of reasons, we clarify some methodological misunderstandings while encouraging further testing of functional groups as biodiversity surrogates.
3Synthesis and applications. We call for caution in making generalizations, and emphasize that case studies on the use of surrogates should be conducted in a systematic manner. This will facilitate robust assessment across studies regarding the usefulness of particular species groups as biodiversity surrogates.
The search for good biodiversity surrogates to facilitate conservation planning is not a new issue (Pearson 1994; Faith & Walker 1996; Andelman & Fagan 2000). Because there is so much biodiversity under threat to be measured and evaluated directly, one often has to rely on biodiversity indicators for designating conservation areas. These indicators or surrogates need to be well known or easy to measure, and should clearly identify which areas are important for biodiversity conservation.
In a recent article in this journal, Sergio and colleagues (Sergio et al. 2006) suggest that top predators may be good biodiversity surrogates, because in their case study on raptors (1) local species richness in raptor sites appears to be greater than in similar sites without raptors, and (2) fewer sites are needed to represent all the species when choosing among raptor sites than among non-raptor sites.
The effectiveness of various surrogate groups for biodiversity conservation is a debated issue. To date, there seems to be no consensus on what taxa or functional groups make good surrogates, what taxa are protected when using surrogates, and how indicator performance is affected by grain, scale or region (Andelman & Fagan 2000; Manne & Williams 2003; Hess et al. 2006). Surrogates are useful if they can be used for conservation planning, when other information about biodiversity is not available. If the choice of protected areas is to be based on the presence of the surrogate(s) only, the question is whether we can really identify important biodiversity areas with such limited information.
As there are different approaches to conservation planning, the performance of a particular surrogate can be assessed in different ways. In this paper, we aim to clarify the use of the concept of surrogates. In particular, we emphasize differences between hotspot and complementarity approaches – concepts that are indirectly addressed by Sergio et al. (2006).
Can top predators be used to detect biodiversity hotspots?
It is important to note, however, that Sergio et al. do not assess whether top predator richness and biodiversity are spatially congruent. Rather they examine whether locations of single top predator occurrences embrace more biodiversity than similar locations without the top predator species, with results that appear to support the use of top predators as a surrogate group for local biodiversity hotspots. This potentially promising finding has to be treated with caution for a number of reasons related to the lack of representativeness of their study survey. For instance, their survey did not provide complete coverage of biodiversity in the area. Instead, for each predator's study area they surveyed the diversity of bird, butterfly and tree species in 1 km2 around 25 raptor-nesting sites, and around control sites (25 locations for spatial controls, and 25 sites for each of the two types of taxonomic controls). The sampling grain differed between groups (e.g. 1 km2 for birds, 20 m2 for butterflies) and, most importantly, the choice of regions sampled was biased towards the area of occurrence of the avian predators. This ignores the possibility that the most important places for conservation could be outside the areas where avian top predators occur.
Even if the results remain true with a more complete survey, a positive relation between top predators and species richness does not necessarily indicate sites of conservation priority; this relationship may be due to raptors preferring areas with habitat gradients or fragmented landscapes, which are often the result of anthropogenic disturbance (Rodriguez-Estrella, Donazar & Hiraldo 1998; Ozaki et al. 2006). In such cases, it might be better to conserve sites based on the presence of other species with more specialized habitat requirements.
Do top predators represent overall biodiversity in reserve-network selection?
The need for efficient and effective allocation of conservation resources has motivated the development of systematic conservation planning tools during recent decades (Lindenmayer, Margules & Botkin 2000; Cabeza & Moilanen 2001). This has been followed by a shift in the search for surrogate groups to identify hotspots of species richness (Prendergast et al. 1993) towards a search for groups of species whose individual distributions would complement one another, with the idea that the same localities would be complementary for non-assessed biotas. Protecting complementary areas for such groups would thereby capture much of the biodiversity of the region as a whole (Manne & Williams 2003). Note the difference with the richness hotspot approach: a complementarity approach does not aim at just selecting the set of sites with the highest number of species, but instead aims at selecting sites that differ greatly in composition and therefore host more species in combination. When searching for surrogates in this context, we are not asking whether the presence of specific species or richness within one group predicts the richness within another group, but whether complementarity within the surrogate group can predict complementarity within another group.
Reliable indicator groups of complementarity have proved difficult to find, with taxon-based surrogate test groups often resulting in no more biodiversity represented than when areas are chosen at random. Figure 1 summarizes the results from several studies. Note that most of the groups that have been examined in more than a single study have mixed results, meaning that they would not be reliable surrogates, despite some positive results.
As in the search for hotspot indicators, complementarity studies have primarily assessed particular taxonomic groups as surrogate candidates; only recently have researchers started to evaluate functional or other groupings of species (Andelman & Fagan 2000; Manne & Williams 2003; Sergio, Newton & Marchesi 2005a; Sergio et al. 2006). Do functional groups hold more promise than taxonomic groups as potential biodiversity surrogates for complementarity-based conservation planning? Sergio and colleagues (Sergio, Newton & Marchesi 2005a; Sergio et al. 2006) consider top predators as a functional group. While we commend the type of study they have undertaken, we have some concerns about the potentially misleading message that one could take from their study. Although Sergio et al. do use the concept of complementarity in one of their analysis, they are not actually testing whether top predators are efficient surrogates in complementarity-based conservation planning. We next describe briefly how to evaluate surrogate performance in such a context, and proceed to clarify what was done in the study of Sergio et al.
Guidelines for the evaluation of surrogate performance in systematic conservation planning
A range of different methodologies has been used to assess the usefulness of various biodiversity indicator groups in systematic conservation planning. Williams et al. (2006) review these approaches and present a robust method to assess complementarity, based on the frequency of false high and false low predictions by the surrogates. Another simple and robust alternative is to assess the coverage of biodiversity in reserve networks selected for the surrogate group. For this latter approach to yield sound results, one should not base the evaluation on a single reserve network for the indicator group, but preferably select a set of solutions (e.g. reserves covering 1, 5, 10 and 20% of the region's area) for which the coverage of other biodiversity is assessed (see, e.g. Williams et al. 2006; Larsen, Bladt & Rahbek 2007).
An important step in the evaluation of surrogate performance is to assess whether the pattern observed could have arisen by chance. In other words, one needs to assess surrogate performance against appropriate null models. We would encourage the comparison of surrogate effectiveness in protecting biodiversity against the following.
2The maximal representation possible for the same area or cost, selecting sites based on information on all species (not only on the surrogate group).
3The effectiveness of other groups, or the effectiveness of using any biodiversity data (for instance groups of species composed of the same number of species selected randomly) when selecting protected areas.
Sergio et al. (2006) conduct a very different type of analysis that cannot adequately answer the question of whether top predators are good indicators of complementarity. Apart from the results being potentially affected by the sampling bias and the choice of algorithm – e.g. a richness-based heuristic instead of a rarity-based, or a stochastic global-search algorithm, both of which yield more optimal solutions (see Margules, Nicholls & Pressey 1988; Cabeza & Moilanen 2001; Lindenmayer et al. 2002) – the main problem is in the way top predator occurrences were used in the complementarity analysis.
First, Sergio et al. (2006) did not apply the reserve selection method to top predator information alone (i.e. minimum number of sites to represent all top predators), and test whether other species of interest would also be included in the optimal minimum set. Instead, they used information on the occurrences of all avian species to select minimum sets of sites either from the 25 sites where top predators occur, or from different sets of control sites, finding that one needs fewer sites when choosing among top predator sites than when choosing among control sites. This result does not tell us how effective raptors would be as surrogates if reserves were selected by using information on these top predators alone. Using information on raptors alone could result in a selection of sites rich in species number but overlapping in composition (i.e. the same species at many sites), while missing sites that might contain non-represented species.
Second, it is important to note that Sergio et al. apply the reserve-selection approach to each study area (corresponding to a different raptor species) separately. Therefore, the analysis undertaken does not assess the complementary value of these different raptor species. Systematic conservation planning aims at the efficient selection of sites with complementary biotas, for which reason they are best applied to heterogeneous regions with turnover in species composition. In the case of Sergio et al., given that the raptor species differed in diet and habitat associations, it would mean including the different study areas in the same analysis, i.e. all top predator localities at once.
Third, Sergio et al. (2006) compare their minimum set selected from 25 predator sites to minimum sets selected from control sites (for spatial and taxonomic controls), but select for each control only a single set of 25 sites. This does not provide adequate evidence to demonstrate that the observed patterns did not arise by chance.
In summary, a useful surrogate group should perform better than selecting the same number of, or equally costly, areas at random, or than using any other group of species as surrogates, e.g. a random combination of the same number of species.
With these comments, and referring to the long list of taxon-based surrogate studies that have given mixed results (Fig. 1), we want to recommend standardization of studies assessing the usefulness of any group as a surrogate for biodiversity in complementarity analyses. Only with systematic studies will we be able to make generalizations and find out whether there is actually a group of species (or environmental variables) that would systematically work across biodiversity levels, grain and scale (Lindenmayer, Margules & Botkin 2000; Hess et al. 2006; Larsen, Bladt & Rahbek 2007), and which features make a particular group into effective surrogate (Manne & Williams 2003; Larsen, Bladt & Rahbek 2007).
Sergio et al.'s (2006) results re-open the debate about biodiversity surrogates, but the use of top predators alone to identify protected areas is not well justified on this basis. While the characteristics of top predators enumerated by Sergio et al. (2006) may make them suitable as indicators of local richness hotspots, top predators may not have the same value as surrogates when aiming at complementary networks. For top predators to be efficient surrogates in the latter context, different top predator species should be associated with different community compositions. To what extent this is true remains to be tested.
We are not saying, of course, that top predators would not be worth protecting, but we emphasize that for surrogates to be reliable in biodiversity conservation their applicability needs to be general and based on systematic analyses. Given the amount of previous evidence suggesting the contrary, a single study in a small region conducted in an unsystematic way cannot be taken as proof of the effectiveness of top predators as biodiversity surrogates. Furthermore, given that a large number of studies suggest that the effectiveness of indicators is context-dependent, we call for caution if planning is based on indicator relationships assessed at different spatial scales or geographical locations.
We thank T. Roslin, I. Hanski, J. Hottola and two anonymous reviewers for useful comments on earlier drafts of the manuscript. This work was supported by the Academy of Finland, grants no. 209017 (M.C.), 202870 (A.v.T.) and LUOVA – Finnish School in Wildlife Biology, Conservation and Management (A.A.).