Regan Early, Department of Biology, University of York, York YO10 5YW, UK (e-mail firstname.lastname@example.org).
1Faced with unpredictable environmental change, conservation managers face the dual challenges of protecting species throughout their ranges and protecting areas where populations are most likely to persist in the long term. The former can be achieved by protecting locally rare species, to the potential detriment of protecting species where they are least endangered and most likely to survive in the long term.
2Using British butterflies as a model system, we compared the efficacy of two methods of identifying persistent areas of species’ distributions: a single-species approach and a new multispecies prioritization tool called ZONATION. This tool identifies priority areas using population dynamic principles (prioritizing areas that contain concentrations of populations of each species) and the reserve selection principle of complementarity.
3ZONATION was generally able to identify the best landscapes for target (i.e. conservation priority) species. This ability was improved by assigning higher numerical weights to target species and implementing a clustering procedure to identify coherent biological management units.
4Weighting British species according to their European rather than UK status substantially increased the protection offered to species at risk throughout Europe. The representation of species that are rare or at risk in the UK, but not in Europe, was not greatly reduced when European weights were used, although some species of UK-only concern were no longer assigned protection inside their best landscapes. The analysis highlights potential consequences of implementing parochial vs. wider-world priorities within a region.
5Synthesis and applications. Wherever possible, reserve planning should incorporate an understanding of population processes to identify areas that are likely to support persistent populations. While the multispecies prioritization tool ZONATION compared favourably to the selection of ‘best’ areas for individual species, a user-defined input of species weights was required to produce satisfactory solutions for long-term conservation. Weighting species can allow international conservation priorities to be incorporated into regional action plans but the potential consequences of any putative solution should always be assessed to ensure that no individual species of local concern will be threatened.
In order to persist in regions of highly fragmented habitat, species often require ensembles of habitat remnants. Compared with single isolates, ensembles potentially increase persistence through increased actual and effective (genetic) population sizes (Joyce & Pullin 2003), recolonization following local extinction, buffering against stochastic and directional environmental change, and reduced likelihood of community-level cascades of extinction (Hanski & Gilpin 1997). Thus the spatial arrangement of protected areas is critical for sustaining viable populations (Cabeza 2003) and should be incorporated into conservation planning (Cabeza & Moilanen 2001).
The first, main, difficulty in incorporating aspects of population viability into large-scale conservation planning is the perceived enormity of the problem. The processes of dispersal and colonization, and how habitat attributes affect demographic processes, cannot be characterized in detail for every species under consideration. Thus conservation planning for multiple species requires short-cut approximations of population viability. ZONATION is a multispecies planning tool (Moilanen et al. 2005) that uses the connectivity of populations within landscapes as a surrogate for viability. Well-connected landscapes contain a concentration of habitat and populations that have the potential to exchange individuals. By prioritizing landscapes based on the population connectivity of the species within it, ZONATION assigns primary importance to the ecological principles that determine persistence. It also retains complementarity (Rodrigues & Gaston 2001) to ensure that all species are well represented within a proposed reserve network.
None the less, any planning approach that aims to protect multiple species in as few areas as possible inevitably offers each species less protection than if conservation was targeted solely at that species (McCarthy, Thompson & Williams 2006). Quantification of the efficacy of multispecies solutions for individual species is required to ensure that solutions are genuinely practical conservation measures and do not threaten some members of the biota. We did this for ZONATION by comparing its results with the highest priority landscapes for individual species.
A second practical consideration of conservation planning tools is that not all species will be considered equally important. Most countries implement some kind of red list approach. Yet while species values are theoretically possible to implement in some reserve selection approaches, this is rarely done in practice (Arponen et al. 2004).
A third practical difficulty is that conservation is usually planned within politically defined regions. Planning conservation within a region, such as a province, state or country, may focus conservation efforts on species that are rare within those regions rather than adequately protecting strongholds of internationally threatened species (Gardenfors 2001). In contrast, purely international planning may overlook local social and economic benefits of biodiversity within each region, as well as policy commitments to maintain national/regional biodiversity (Gardenfors et al. 2001).
The three goals of our work were to: (i) evaluate whether a multispecies reserve selection tool, ZONATION, represents a practical solution for the long-term conservation of individual species; (ii) investigate whether species weights can be used to achieve individual species targets in multispecies analyses; (iii) explore how weightings can be used to enforce large-scale conservation priorities at a local scale. Analyses were carried out on the well-characterized British butterfly fauna (Asher et al. 2001).
creating connectivity surfaces
Each species’ distribution was used to produce a grid-based ‘connectivity surface’. For any species, the connectivity of a given cell represents the likelihood that an individual will disperse to it. This is dependent on the location, frequency and size of populations in the surrounding area. The probability that a dispersing individual will arrive at grid cell i from all source populations k was estimated with the equation:
( eqn 1)
where k represents all cells other than i in a study area, dik is the distance between cells i and k, α is a species-specific dispersal parameter determining the relationship between the number of dispersers and distance travelled (Hanski, Kuussaari & Nieminen 1994) and Ak is a measure of the size of the source population. Presence–absence data can be used but abundance data improve accuracy in the identification of good-quality landscapes.
For any species, the connectivity score of a region is the sum of connectivity of all grid cells in it. The probability of persistence of a given species is expected to be highest in regions with the highest connectivity scores (Moilanen et al. 2005).
zonation: landscape prioritization
ZONATION selects priority landscapes based on input surfaces for each of a set of species. Each surface is processed simultaneously, and cells within the study region are iteratively discarded; the least-valued cells are discarded earliest. The output is the ranking of cells in order of removal, allowing identification of the most important areas for butterfly persistence when a certain proportion of the land surface remains. The following rules were used: (i) cells (here 1 km2) can only be removed from the edge of the remaining area, i.e. adjacent to a cell that has already been removed, or from the boundary of the landscape; (ii) at each step, the cell that contains the smallest fraction of any species’ total remaining connectivity is removed. This fraction is calculated by:
(proportion of remaining species’ connectivity in cell × species weight) ÷ proportion of species’ original distribution that is retained in the remaining cells( eqn 2)
Cells at the core of a species’ distribution are more highly valued than marginal ones because they contain a larger proportion of the species’ total connectivity, so will have a large solution to equation 2. For the same reason, cells within the distribution of a rare species are more highly valued than those in the distribution of a common species.
After every step, equation 2 is recalculated for all cells eligible for removal. As a species’ distribution is degraded, the solution to equation 2 increases. This ensures that part of every species’ distribution will be preserved in the last few remaining cells. Conducting ZONATION on connectivity surfaces also selects areas that will retain the structural integrity of landscapes.
management unit recognition
Conservation decisions would be helped if we knew which areas were parts of the same ecological system and required coherent management. To this end, we grouped ZONATION-prioritized cells into management units based on the faunal dissimilarity and distance between clusters of cells (Moilanen et al. 2005).
First, clusters of cells could only be retained in the final solution if they represented the very best areas; they were required to contain one or more cells that were in the top-ranked 0·5% of cells. Secondly, two or more clusters could be considered as part of a single management unit if they were close enough to each other to be linked by occasional, long-distance dispersal events (within 10 km). Finally, clusters were only grouped if they shared a similar fauna. Species’ abundances were compared between clusters. If the abundances of fewer than three species differed by one order of magnitude or more, then clusters were considered to be part of the same management unit.
Throughout 1995–99, sightings of British butterfly species were collated by Butterfly Conservation (a UK-based society working to protect butterflies and moths) for the Butterflies for the New Millennium project (Asher et al. 2001). The study system comprises 57 butterfly species: 54 are breeding residents in Great Britain and three are regular migrants (see Appendix S1 in the supplementary material). We did not consider two reintroduced species, Maculinea arion (Linnaeus 1758) and Lycaena dispar (Haworth 1803), the priority areas for which depend more on the potential for habitat restoration than on their current distributions. Records were converted to three measures of abundance (Ak): one individual, 2–9 individuals and 9+ individuals sighted on a single-site visit. The data were used at a 1-km grid cell resolution. For cells in which the species was recorded on multiple occasions, the largest Ak value in each grid cell was used. Because the recordings were not fully standardized, these population sizes of each grid cell (Ak) were treated as ordinal categories, and scored as 1, 2 and 3, respectively. If connectivity is calculated using accurate population sizes, Si (equation 1) represents the expected average number of immigrants to a grid cell within a year. However, because Ak is only a representation of population size, the connectivity of any grid cell here is a relative index of expected immigration rate. This is what is summed to obtain the ‘connectivity score’ of a group of cells. Approximately 1·55 million species location records were included.
Species were grouped into broad dispersal categories following the ranking of dispersal ability by Cowley et al. (2000). Groups were allocated appropriate α-values (equation 1) based on the proportion of individuals expected to move certain distances. Seven groups of species were recognized, with α-values of 3, 2·5, 2, 1·75, 1·25, 0·5 and 0·25 (see Appendix S1 in the supplementary material). However, lower α-values (higher dispersal) would better capture the longer term connections between landscape elements that could be vital to the persistence of populations. Therefore, based on a detailed study of the long-term dynamics of one of the species considered here, Euphydryas aurinia (Rottemburg, 1775) (R. Early, unpublished data), we halved the α-values assigned to each species.
high-priority landscapes for individual species
Butterfly Conservation lists 23 species as high and medium priority in the UK (Warren et al. 1997). We used connectivity surfaces to identify the ‘functional landscape(s)’ predicted to be most important for each of these species. A functional landscape for a species is a region in which all habitat fragments are connected in some way, and which is predicted to operate as one dynamic system over a long period of time. In the absence of further habitat alteration, it can be argued that the largest and best connected functional landscape is the region where a species is most likely to persist.
Connectivity surfaces were transformed into GIS coverages using ArcGIS v9 (ESRI, Redlands, California). For each species, every cell with a connectivity value above a threshold was selected to produce clusters of cells, each of which is a functional landscape. The ‘best’ landscape for any species was that with the greatest summed connectivity score. We investigated the impacts of choice of connectivity threshold using values of 0·0001, 0·001, 0·01 and 0·1. A threshold of 0·1 defines a boundary expected to be reached by 10% of individuals originating from a single source subpopulation. A very strongly dispersing species might be given an α-value of 0·25 (in our connectivity calculations this value would be halved to account for long-term population processes) for which the distance reached by 10% of source population individuals is 9·2 km (equation 1). For relatively sedentary species given an α-value of 3, this becomes 0·8 km. The lowest threshold of 0·0001 defines a boundary expected to be reached by 0·01% of individuals. The corresponding distances for a highly mobile and a sessile species are 36·8 km and 3·1 km, respectively. Therefore, the size of the landscapes allocated to populations is large when threshold values are small.
If a rare species exists in small populations, even its best single functional landscape may be prone to extinction. Therefore, we determined a minimum connectivity score that must be attained for each species. Melitaea cinxia (Linnaeus 1758) is the most range-restricted butterfly species in Britain, occurring predominantly in a small part of the Isle of Wight. Its populations may be assumed to have a high likelihood of persistence as it is apparently stable in its current distribution (Asher et al. 2001). The connectivity score of its entire British distribution is 48·8. Thus, target functional landscapes for all other species were required to exceed this. In practice, the target functional landscape of almost all species contained more than twice this connectivity. For the few species that did not, additional functional landscapes were prioritized until the required connectivity score was exceeded.
implementing species weights within zonation
We implemented the following species weighting schemes in ZONATION. The weights applied to each species are contained in Appendix S1 (see the supplementary material).
National species weights
Butterfly Conservation classifies species as low, medium and high priority in the UK (Avery et al. 1994; Warren et al. 1997). We created a weighting scheme called UK1,5,10, in which low-, medium- and high-priority species were given weights of 1, 5 and 10, respectively. To evaluate how the numerical scale of weights affects landscape prioritization, we also created a weighting scheme called UK1,2,3 in which we assigned weights of 1, 2 and 3 to low-, medium- and high-priority species, respectively. Seven species were given high weights and 16 were given medium weights.
European species weights
The Red List produced by the Council for Europe has assimilated national lists of threatened species into a Europe-wide risk assessment (Van Swaay & Warren 1999). Species not classified at risk were assigned a weight of 1, ‘lower risk (near threatened)’ species were assigned a weight of 5 and ‘vulnerable’ species a weight of 10. This weighting scheme was called CE1,5,10. Again, weights of 1, 2 and 3 (3 given to the highest priority species) were also used and this scheme was called CE1,2,3. Three species were given high and two species were given medium weights.
The connectivity maintained within every target landscape and ZONATION solution was calculated as the sum of connectivity values of each grid cell within them for all species (equation 1). To compare ZONATION and target landscape reserve networks, we calculated ZONATION solutions of equal sizes to the four target landscape solutions.
high-priority landscapes for individual species
Following the delineation of target landscapes according to connectivity thresholds, some of the 23 UK high- and medium-weight species required multiple target landscapes to fulfil our criteria for a high likelihood of population persistence (the target landscape(s) for each species was required to exceed the connectivity score of the whole distribution of M. cinxia). Using the highest connectivity threshold of 0·1, the single best functional landscapes defined for a quarter of the target species did not attain the criteria: four species required two landscapes and two species required three landscapes. At thresholds of 0·01, 0·001 and 0·0001, three, two and two species, respectively, required multiple functional landscapes to attain the target.
The target landscapes delineated for the 23 species using the four connectivity thresholds (0·1, 0·01, 0·001 and 0·0001) resulted in the prioritization of 3%, 12%, 30% and 47% of Britain's land surface. The average total connectivity of the 23 focal species retained in these landscapes was 46%, 78%, 93% and 97%, respectively. Within the cells that were selected, the actual area used by the populations will be far smaller (Cowley et al. 1999). Target landscapes for the different thresholds were largely nested, indicating that most British butterfly species do have a distinct ‘best landscape’. Eight-two per cent of the 0·1 connectivity threshold solution was contained within all three other solutions, 75% of the 0·01 threshold solution was contained within both the 0·001 and 0·0001 threshold solutions, and 100% of the 0·001 threshold solution was contained within the 0·0001 threshold solution.
The relationships between the connectivity included in the target landscape solutions and the proportion of Britain they include are convex curves (Fig. 1). The curves are very shallow for low-weight species. Because low-weight species are widespread, their protection is largely determined by the total area of reserve designated. Even though the proportion of the connectivity retained for these species was relatively low, the absolute amount of connectivity retained was far greater than for the high- and medium-weight species.
target landscape protection by zonation
ZONATION's multispecies approach almost always retained more of a species’ original connectivity than the ‘best’ (target) functional landscape(s) (Fig. 1). This was because marginal, low connectivity, cells from target landscapes were discarded in favour of the core parts of lower-ranked landscapes.
A comparison between the target landscape and ZONATION approaches is shown in Fig. 2. ZONATION splits the target landscapes into smaller subsections. This results in a reserve network that identifies the same ‘hotspots’ as the single-species approach and that is also more extensive. However, the fragmentation of the target landscapes by ZONATION becomes more pronounced and problematic as the solution area decreases. The more range-restricted a species, the less severely ZONATION fragments their target landscapes. For example, even in the smallest solution (3% of land surface), both of the functional landscapes targeted for Melitaea athalia (Rottemburg 1775) (the second rarest British butterfly species) were fully included in ZONATION's reserve network.
weighting species within zonation according to national importance
ZONATION weighted according to the UK1,5,10 scheme protects slightly more of the connectivity of the UK high- and medium-weight species than unweighted ZONATION (Fig. 1). The impacts of weighting ZONATION were investigated further by comparing how well unweighted and UK1,5,10-weighted ZONATION solutions protect the target landscapes. We took ZONATION solutions that retained the same number of cells as each of the target landscape approaches using the four different connectivity thresholds. For each UK high- and medium-weight species, we calculated the proportion of its target landscape that was contained within the corresponding ZONATION solution (Table 1).
Table 1. Average connectivity of each species’ target landscape contained within two ZONATION solutions: unweighted and weighted according to the UK1,5,10 scheme. At each connectivity threshold, the areas of land surface covered by both ZONATION solutions and the target landscape solution are identical
Target landscape connectivity threshold
Percentage land surface of Britain remaining
Weighting scheme applied to ZONATION
Mean proportion of the connectivity of individual species’ target landscapes maintained within the ZONATION solution
UK high- weight species
UK medium- weight species
The UK high- and medium-weight species are those for which the target landscapes were designed. Therefore, applying UK1,5,10 weightings focuses ZONATION to maintain more of the target landscapes, and reduces their fragmentation. Note that, in the smallest solution (3% of Britain's land surface), applying UK weights actually performs worse than applying no weights for the medium-weight species. In this very tiny reserve network, the weightings favour the high-weight species to the detriment of the medium-weight species.
The reduced fragmentation of the best areas for highly weighted species can be useful. Carterocephalus palaemon (Pallas 1771) is a UK high-weight species whose entire British distribution is restricted to one region of western Scotland, which can be split into northern (C. palaemon, i) and southern areas (C. palaemon, ii; Fig. 4c). At the 0·01 connectivity threshold, the C. palaemon target landscape protects just the northern area (Fig. 4c). The corresponding UK1,5,10-weighted ZONATION solution protects more of this region, in fewer, larger sections (representing the best connected/most densely populated parts of the landscape), than unweighted ZONATION. It also maintains the second-most, and only other, important area for this butterfly in Britain (shown in brown; Fig. 4d).
comparing national and european weighting schemes
Figure 4 shows the areas selected in Britain by the national and European priority-weighted solutions after ‘management unit recognition’ was performed on ZONATION's last remaining 10%. Each management unit is shown as a single colour. The grouped areas cover 7% of the country, the remaining 3% being small areas that do not contain areas that are critical to conservation (using a final 0·5% selection criterion; see the Methods). Interestingly, where ZONATION has fragmented target landscapes, management unit recognition tends to ‘put them back together’, i.e. group habitat fragments into clusters according to whether they belong to a target landscape.
Following the management unit recognition procedure, the similarity between the UK1,5,10 (Fig. 4a) and European CE1,5,10 (Fig. 4b) solutions is 58% (the similarity between the pre-recognition solutions is shown in Table 2). The differences are predominantly caused by alterations to the edges of selected priority areas, although a few completely different small areas are selected under either scheme. Some large landscapes selected under UK weightings are reduced in size or fragmented under European prioritization. The consistent selection of the cores of priority areas is illustrated by the fact that the amount of connectivity protected for the 23 target species is very similar under both UK1,5,10 and CE1,5,10 weights (Fig. 1).
Table 2. Proportions of ZONATION-selected priority areas (before management-unit recognition) that are identical under different weighting schemes, when 10% of Britain remains in the solution
Irrespective of how much land surface remains, UK high- and medium-weight species are slightly better protected under UK than European weightings on average (Fig. 1). All five European high- and medium-weight species are better protected under CE weightings than UK weightings. The CE1,5,10 weights reduce the amount of area selected in southern England and increase the amount selected in Scotland. This is because of three European high- and medium-weight species that have a large proportion of their British distribution in Scotland. Applying CE weights shifts their protection from the margins to the cores of their range. This shift is marked because Scotland is relatively depauperate in butterfly diversity; few species can be protected by any one priority area and to increase protection of a few species requires a disproportional amount of area. However, the shift succeeds in increasing the representation of important European species. In the most extreme case, the top 2% of the ZONATION solution retains 26% of the initial connectivity of Erebia aethiops (Esper, 1777) under CE weights rather than 6% under UK weights.
As with the UK weightings, we assessed how European weightings might affect the conservation of viable populations by comparing them with the target landscape solutions (Fig. 3). Target landscapes of important European species were better protected under CE than UK weights (e.g. Coenonympha tullia (Muller, 1764); Fig. 4). Target landscapes of species without European importance were given less protection.
numerical scale of weighting scheme
Table 2 shows the overlaps between all weighted and unweighted ZONATION solutions when 10% of Britain remains in the solution. These are the differences in ZONATION-selected cells before management unit recognition is performed. Using both UK and European weighting schemes, the ‘1, 2, 3’-weighted solutions are intermediate to the unweighted and ‘1, 5, 10’ solutions. There was more similarity between the UK1,2,3- and CE1,2,3-weighted solutions (83% overlap) than between the UK1,5,10- and CE1,5,10-weighted solutions (66% overlap). The CE1,5,10-weighted solution had least similarity to the others; under this scheme only five species were given a weight of more than one.
high-priority landscapes for individual species
The protection of species in small isolated reserves has only had limited success in maintaining viable populations (Warren 1993; Rodrigues 2000). When existing landscapes are already highly fragmented, it is critical to be able to identify clusters of habitat fragments that are likely to allow species to persist indefinitely. We identified such clusters for endangered British butterflies by identifying target landscapes based on connectivity surfaces. The connectivity metric used (equation 1) is dependent on the size of source populations and the dispersal characteristics of individual species (α-value); it has been shown to predict colonization events well (Moilanen & Hanski 2001). Thus, while species disperse in a variety of different ways (Matter, Roslin & Roland 2005), as long as α-values can be estimated equation 1 can still be applied. However, if species’ dispersal patterns are believed to differ strongly from the kernel used here, the metric could be replaced by a more appropriate measurement.
The capacity of a landscape to support a viable population often has a threshold (Hanski, Moilanen & Gyllenberg 1996). Thus, while many reserve design approaches attempt to prioritize persistent areas of species’ distributions (Rothley 1999; Possingham, Ball & Andelmann 2000; Briers 2002; McCarthy, Thompson & Possingham 2005), they do not make explicit assessments of the quality of the underlying landscape. This makes it difficult to judge where the threshold for persistence lies. The size of landscapes required to support a viable population depends on our expectations of species’ dynamics (Cabeza 2003). Large landscapes are necessary if subpopulations are extinction-prone (Baguette & Schtickzelle 2003) or if occasional long-distance dispersal events link distant populations and are important for persistence (Baguette 2003). Smaller landscapes will be efficient if dynamics are stable, and persistence may be dependent on the behaviour of central well-connected subpopulations (Wilson et al. 2002).
However, these statements are qualitative. The ability of any approach to protect viable populations can only be diagnosed by detailed population viability analyses. Unfortunately these have rarely been conducted on the sort of multispecies systems on which reserve selection is needed, as they require detailed data on each species’ population processes (Carroll et al. 2003). Therefore, it is uncertain which of the threshold values we used to identify single-species target landscapes would be most appropriate. To provide some guidance, we compared functional landscapes defined by our connectivity thresholds with the known distributions of populations and habitats of Aricia agestis (Denis & Schiffermüller, 1775) in north Wales (Wilson et al. 2002). Landscapes delimited by a threshold of 0·1 contained the core of the populations but excluded some habitat that contributed to the persistence of the populations, as simulated over 100 years. Landscapes delimited by a threshold of 0·01 included complete networks. Lower thresholds did not add any further useful habitats. Thus, functional landscapes defined by a connectivity threshold of 0·01 appeared to generate appropriately sized units for conservation. Although appropriate threshold values may vary among taxa, connectivity surfaces do define relative likelihood of persistence in a reliable manner.
does zonation represent a practical solution for long-term conservation of individual species?
While protecting viable populations is paramount, solely protecting an endangered species’ strongest population could lead to the extinction of potentially viable populations elsewhere and reduced protection for lower priority species. ZONATION maintained good representation of the UK high- and medium-weight species both within their target landscapes and throughout their range. Thus it produces a compromise between the identification of areas that contain viable populations and the complementary minimum-set reserve selection algorithms that identify a network that includes multiple locations for all species (Margules, Nicholls & Pressey 1988). As solution sizes become smaller, there is a risk that the remaining ‘protected areas’ will no longer support functioning networks of populations that are viable in the long term: ZONATION protected least of the target landscapes at the smallest solution size (3% of Britain). We suggest that ZONATION reserve designs (and those of other multispecies reserve selection approaches) should always be investigated for robustness to population extinction, particularly when a low percentage of a study area is selected. None the less, because connectivity is closely associated with population concentration and likelihood of persistence, ZONATION reserve quality may be easier to interpret than when, for example, reserve aggregation is achieved through boundary penalties (Possingham, Ball & Andelmann 2000).
can species weights be used to achieve individual species targets in multispecies analyses?
We applied national species values (Warren et al. 1997) to the selection of priority areas. Weighting important species within ZONATION improved their protection over unweighted ZONATION, both throughout Britain and within their target landscapes. Provided that categories of species of conservation concern can be recognized, weighting them increases the ability of ZONATION to identify a reserve network containing viable populations of all species. In the case of C. palaemon, ZONATION weighted according to UK species targets (UK1,5,10) is clearly a more appropriate conservation strategy than protecting the species solely within its target landscape.
ZONATION's ‘management unit recognition’ feature grouped clusters of prioritized cells into conservation units that closely matched the single-species target landscapes. Because fragments are grouped according to similarity in their entire fauna, this was actually more able than the single-species approach to identify a landscape that is a true biological and management entity. While further work is needed, this may be a way of bridging the gap between a species-centric definition of conservation units and strategies based, for example, on the resilience of biotic communities (Kemper, Cowling & Richardson 1999) or their unique characteristics (Turner et al. 1999).
We conclude that the benefits of multispecies analyses such as ZONATION are important to the effective conservation of biodiversity. Weighting species within them causes their solutions to achieve conservation goals of increased representation and viability for individual threatened species.
how should weightings be used to implement international priorities at a local scale?
If each country in Europe prioritized conservation according to species rarity or threat within its borders, landscapes that offer species the greatest chance of long-term persistence could be overlooked if they occur in countries where the species is not yet recognized as threatened. But conservation is usually planned at a national scale. We investigated whether using European species’ targets (CE weights) to design a British reserve network is an appropriate method of conserving international species, and how this would affect species that are regarded as threatened only within Britain.
For species that are rare in the UK but widespread in Europe, applying CE rather than UK weights did not greatly reduce their protection. This is because ZONATION is strongly influenced by the distributions of the input species; it maintains a high proportion of the ranges of localized species. However, some of these species were offered little protection within their ‘best’ British landscapes. This is not a trivial difference. Small alterations in the extent of reserves can have a major effect on survival probability because of the threshold nature of landscape capacity to support a population (Hanski & Ovaskainen 2000).
Given that the quality of data on threats and decline is higher in Britain than throughout Europe, it may be unwise to base all weighting decisions on the latter. In any case, the UK weights used incorporate an element of the European importance of national populations (Warren et al. 1997). Thus increasing the weights of species in Britain that are threatened in Europe may be more appropriate than unadjusted adoption of the European-scale weights. The unadjusted European-scale priorities might be more appropriate for a country with a larger number of European priority species.
Weighting schemes are the outcome of how groups of species are identified and the relative values they are given. The interplay between these two elements must be considered thoroughly because priority areas are designated for highly weighted species at the expense of lower weighted species. These effects were greatest when (i) there was a large difference between the highest and lowest species weights (i.e. the 1,5,10, scheme); (ii) few species were weighted highly (i.e. the CE scheme); and (iii) when a small proportion of the original area was to be prioritized (i.e. the final 2% of ZONATION's solution). Applying a smaller numerical range of weights (1,2,3) to the European and UK groups of species gave very similar solutions. These weights may not be strong enough to identify where protected areas would be important for highly weighted species. More extreme weighting systems than our 1,5,10 scheme (say 1,10,100) would produce solutions that effectively conserve the top-weighted species but largely ignore the others.
On a more general note, there will be both uncertainty and controversy in the structure of any species list (Possingham et al. 2002; Noss 2004; Robbirt, Roberts & Hawkins 2006) as well as its application (Pearman 2002). When there are multiple criteria for assigning values to species (e.g. global vs. national concerns, economics, genetic uniqueness, ecosystem contribution and cultural importance), we recommend that they should be implemented in conservation planning separately and the results compared. This allows us to understand what will be the impact of accepting any of a set of values and can feed into other aspects of conservation planning. For example, are priority areas defined using any of the species valuations facing particular threats? Do any of the species valuations lead to fundamentally different prioritization? Thus, with careful application, reserve design tools can be made to achieve their long-awaited integration into practical conservation planning (Prendergast & Nicholls 1999).
While it is desirable to maintain multiple landscapes for each species, there also is a strong argument for maintaining intact their most valuable landscape(s). Defining target functional landscapes according to connectivity surfaces and thresholds enables us to judge how effectively the priority areas arrived at by multispecies reserve selection may protect viable populations. Ninety-nine per cent of the target landscapes of UK high-weight species and 90% of those of UK medium-weight species could be retained in the priority areas selected by UK1,5,10-weighted ZONATION solution covering 12% of Britain's land surface. We preferred UK-weighted ZONATION to the single-species approach because, as well as protecting most of the target landscapes, on average ZONATION protected over 20% more of each species’ connectivity in other core areas.
By differentially weighting species, we can make small or large changes to a reserve network design, depending on the numerical scale of the weights used. If defined consistently for all species according to reasonable criteria, weights can manipulate reserve designs toward the needs of highly weighted species. ZONATION allows this to occur without substantial damage to lower weighted species because it responds to species’ distributions more than to the species weights used here. When conservation is planned at a local scale, weighting species according to their global status can help prioritize areas within that region that are of greatest international importance for the species. Because the protection of locally rare species may be reduced to some extent, we recommend simultaneously identifying the ‘best landscapes’ for each species in order to evaluate these impacts.
Finally, the output of any reserve selection procedure is not the final ‘on-the-ground’ solution. Within proposed priority areas, habitats and populations may cover only a small fraction of the land surface. Recognizing that habitat fragments within a landscape are functionally connected represents the first step in the development of a conservation plan. Detailed analysis of the potential performance of a reserve for individual species, as well as the species assemblage, should follow wherever possible. This may include identification of critical habitat fragments and high-quality intervening habitat, and a more functional characterization of species’ dispersal patterns. However, the approaches defined here make a substantial step towards delivering a practical approach that can actually be adopted in conservation planning. We believe that it should be possible to retain viable populations of most, if not all, butterflies in Britain in the landscapes we have identified.
We are very grateful to the many recorders who contributed to the Millennium Atlas data set and to Richard Fox and Butterfly Conservation for making these data available to us. Also to Atte Moilanen for allowing us access to ZONATION and being so helpful. A. Dumbrell, A. Franco and J. Hill kindly commented on the manuscript. The study was made possible by funding from the Countryside Council for Wales.