Introduction
In most developing countries the human population increase of the past 50 years has been linked with a significant increase in natural resource exploitation (Naughton-Treves et al., Reference Naughton-Treves, Kammen and Chapman2007) and this has been most severe for wood (Luoga et al., Reference Luoga, Witkowski and Balkwill2000). Although protected areas represent an appropriate way to conserve natural resources they are usually inadequate in size and do not cover the full range of features in need of conservation (Pressey et al., Reference Pressey, Cowling and Rouget2003). Consequently, there is a need to conserve resources outside formally protected areas by creating a network of formal conserved areas interlinked by a matrix that creates corridors for plant and animal dispersal (Rouget et al., Reference Rouget, Cowling, Lombard, Knight and Kerley2006; Smith et al., Reference Smith, Goodman and Matthews2006).
One of the key issues in systematic conservation planning is to determine how much needs to be protected (Eeley et al., Reference Eeley, Lawes and Reyers2001; Sanderson et al., Reference Sanderson, Redford, Vedder, Coppolillo and Ward2002). Setting targets has been the crux of conservation planning since the 10% global landmass target proposed at the 3rd World Congress on National Parks in 1982 (Pressey et al., Reference Pressey, Cowling and Rouget2003; Tear et al., Reference Tear, Kareiva, Angermeier, Comer, Czech and Kautz2005). However, such targets have promoted the protection of land irrespective of its value for conservation.
Setting targets is currently achieved by integrating the sum of individual targets of all landscape features (soils, topography, climate, plants, animals) to derive a regional target (Pressey et al., Reference Pressey, Cowling and Rouget2003). Regional targets are then used to assess the effectiveness of current conservation areas and develop new management plans. The most severe limitation for setting such targets lies in the lack of biodiversity data, which in some cases could imply that short of conserving 100% of a region, conservation will fail its purpose (Pressey et al., Reference Pressey, Cowling and Rouget2003). Moreover, targets should preferably describe quantitative aspects, as they are more tangible and facilitate further use and refinements, rather than qualitative aspects. Quantitative targets are constructed from evidence, and are therefore defensible.
Setting broad-scale conservation targets requires reviewing smaller scale targets, and assimilating these with conservation planning software (Desmet & Cowling, Reference Desmet and Cowling2004). Such software exists for the selection of priority areas for conservation but tends to be most effective at a broad geographical scale (Eeley et al., Reference Eeley, Lawes and Reyers2001; Reyers et al., Reference Reyers, Fairbanks, Van Jaarsveld and Thompson2001; Pressey et al., Reference Pressey, Cowling and Rouget2003). Such regional overviews are based on a review of local expert judgement within the region of concern (Cowling et al., Reference Cowling, Pressey, Sims-Castley, Le Roux, Baard, Burgers and Palmer2003). However, local experts seldom have the time or budget for the detailed, long-term population viability analyses and habitat modelling (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001; Desmet & Cowling, Reference Desmet and Cowling2004) required for setting scientifically and ecologically acceptable targets at the local scale. As a consequence, expert opinion at the local scale is usually more qualitative than quantitative. Methods that can assist establishment of quantitative conservation targets at the local level are therefore indispensable. Such methods should be simple, easy to use, sufficiently adaptable to incorporate new data, and should work with a limited amount of relevant information such as is typically gathered from baseline ecological studies. Such studies are often the only data available to conservation authorities (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001).
Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) developed such a method for setting conservation targets for plant species when there is insufficient data and time. They applied their method to rare plant species because such species usually require a specific conservation action. In this study we expand this method, evaluating its applicability for landscape species, where landscape species are a derivative of umbrella species (Sanderson et al., Reference Sanderson, Redford, Vedder, Coppolillo and Ward2002). Landscape species are defined as biological entities occurring over large, ecologically diverse areas and having a significant impact on structure and function of local natural ecosystems within a timespan applicable to human management targets (Sanderson et al., Reference Sanderson, Redford, Vedder, Coppolillo and Ward2002). Through conservation of landscape species the conservation of the whole landscape and the co-occurring species could potentially be achieved.
Conserving umbrella species based on a prospective selection of occurrence patterns and ecological traits (Fleishman et al., Reference Fleishman, Murphy and Blair2001) may be a good short-cut for conserving co-occurring species. However, umbrella species are often poorly selected and results may therefore fall short of expectations. Selection based on conservation status, failure to assess co-occurrence of similar taxonomic groups and species richness across the target landscape, and selection of habitat generalists are common pitfalls in umbrella species selection. The amount of information required to select an umbrella species properly may render it anything but a short-cut (Seddon & Leech, Reference Seddon and Leech2008). The ideal umbrella species should be neither rare nor ubiquitous, and offer a significant chance of co-occurrence with other species (Fleishman et al., Reference Fleishman, Murphy and Blair2001). A recent revision of the umbrella species concept narrowed the selection criteria to seven from 17 in an attempt to provide a simpler concept (Seddon & Leech, Reference Seddon and Leech2008). These seven criteria are: well-known biology, large home range, high probability of population persistence, co-occurrence of species of conservation interest, management needs not specifically beneficial only to the umbrella species, moderate sensitivity to human disturbance, and ease of monitoring. This revision is therefore similar to the broader concept of landscape species that rely on five criteria: area requirements, heterogeneity in habitat utilization, ecological functionality, vulnerability to human land-uses and socio-economic significance. Therefore, by selecting appropriate landscape plant species we believe that most co-occurring common and rare plant species will be included, and to some extent the animals that depend upon them, which may make the conservation of these landscape plants an effective tool for conservation purposes.
By adapting and applying the method of Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) over a geographical region identified by systematic conservation planning as important for biodiversity conservation (Eeley et al., Reference Eeley, Lawes and Reyers2001; Smith et al., Reference Smith, Goodman and Matthews2006), our aim is to determine the minimum area of each vegetation community that needs to be fully conserved in a network of conservation areas.
Study area
The study area lies in the Maputaland region of South Africa, with two existing conservation sites: Tembe Elephant Park (30,000 ha) and Tshanini Community Conservation Area (2,420 ha) in the Manqakulane rural community. Maputaland is the northern tip of the Maputaland–Pondoland–Albany biodiversity hotspot (Smith et al., Reference Smith, Goodman and Matthews2006). It consists of a sandy plain interspersed with ancient littoral dunes that lie north–south, covered by a mosaic of Sparse to Closed Woodland with patches of Sand Forest (Matthews et al., Reference Matthews, Van Wyk, Van Rooyen and Botha2001; Gaugris et al., Reference Gaugris, Matthews, Van Rooyen and Bothma2004; Gaugris, Reference Gaugris2008).
The four main vegetation types are: Sand Forest, and Closed, Open and Sparse Woodland (Gaugris, Reference Gaugris2008; Gaugris & Van Rooyen, Reference Gaugris and Van Rooyen2008), the latter best described as a wooded grassland. Sand Forest is biologically the most diverse vegetation type in Maputaland and one of the most important in the Maputaland–Pondoland–Albany hotspot (Matthews, Reference Matthews2006). However, Maputaland’s woodlands are nearly as species rich as the Sand Forest (Matthews, Reference Matthews2006). Vegetation dynamics in the region are not fully understood but they may be driven by relatively frequent small-scale disturbances from incursions by large mammals (Gaugris, Reference Gaugris2008). Large-scale disturbances induced by increasing human populations outside conserved areas and high densities of large mammals inside conserved areas are challenging the resilience of both forest and woodland ecosystems (Botes et al., Reference Botes, McGeoch and Van Rensburg2006; Gaugris, Reference Gaugris2008).
Methods
While keeping to the same terminology, we used and adapted the method of Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) to suit our approach and the conditions of our case study. This 12-step method (Fig. 1) provides a framework within which knowledge of each selected species can be evaluated, thereby facilitating a discussion about how best to calculate the area required to protect groups of associated species. Apart from changes described immediately below, the modifications to the original method are described in our description of the 12 steps. We have tried to keep the methodology as simple as possible to make it accessible.
The process flow chart of the 12 steps in the method of Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001). The method is a combination of coarse (large scale for the regional level) and fine (local scale) considerations, with qualitative and quantitative inputs. The method should ideally be conducted by a panel of experts or people with local knowledge, who guide the decision process and assess the various risks considered. Step 12 is the bridge between qualitative and quantitative aspects.
One of the main modifications is with regards to disturbance region, which we define to be a fixed area of a size that is susceptible to be fully disturbed by at least one disturbance within the time frame of that disturbance and subject to the same decision-making governance. We thus set the size of a disturbance region to 3,600 ha, which is approximately half the area of a rural community’s tribal land in Maputaland (Peteers, Reference Peteers2005). This size is also slightly less than half of the area for reliable representation of meteorological conditions around a meteorological station (Yeh et al., Reference Yeh, Wensel and Turnblom2000). These two considerations ensured that each region could be considered homogeneous with respect to human-related disturbance or localized weather events, thereby incorporating both human and local weather variability within their respective time frames.
We view uncertainty differently to Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001), who considered that high key (all is well) and low key (all is bad) analyses should be conducted in parallel to provide an uncertainty range reminiscent of interval arithmetic. To do this they noted that in most cases these high and low keys will have to be set subjectively. In our approach we evaluated the minimum viable population size (F, see definition below) in an area of 36 × 42 km divided into a grid of 42 square disturbance regions of 3,600 ha, and assumed that the variation within the 42 regions provided a reasonable estimate of uncertainty, with upper and lower bounds based on evaluated threats, thus avoiding running parallel supplementary analyses. In effect, we applied the method to each disturbance region in the area, as Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) suggest. This approach to uncertainty was justified by the fact that most of the 42 regions have a similar ecological potential (Gaugris et al., Reference Gaugris, Matthews, Van Rooyen and Bothma2004) but are subjected to different land tenure and land use.
Prerequisites Our method requires the selection of one or two plant species that are characteristic of a particular vegetation type. The species must represent the vegetation type in such a way that their conservation ensures the conservation of the vegetation type as a whole, including its dynamics (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001). Our choice of plant species followed recommendations for landscape species selection (Sanderson et al., Reference Sanderson, Redford, Vedder, Coppolillo and Ward2002). Cleistanthus schlechteri (Pax) Hutch. and Newtonia hildebrandtii (Vatke) Torre were chosen as landscape tree species of the Sand Forest. Both grow into large canopy trees, are long-lived, have a wide geographical range in southern Africa (Pooley, Reference Pooley1997), and are used for firewood and building material (Gaugris et al., Reference Gaugris, Van Rooyen and Van der Linde2007). Hymenocardia ulmoides Oliv. and Sclerocarya birrea (A. Rich.) Hochst. were the landscape tree species chosen for the Woodlands. H. ulmoides is an abundant species, geographically widespread in KwaZulu-Natal, and it provides good building material (Pooley, Reference Pooley1997). S. birrea is an important food tree in southern Africa and is geographically widespread on the subcontinent (Emanuel et al., Reference Emanuel, Shackleton and Baxter2005).
Step 1 The method firstly requires the establishment of F, the minimum viable population size, defined as ‘the population size that faces a 0.1% probability of falling below 50 adults at least once in the next 50 years, assuming no detrimental human effects’ (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001). This represents a quasi extinction risk and provides a background risk against which it is possible to measure the utility of conservation actions. The 50 years benchmark reflects concerns for risks over which current management decisions may be effective. In the original method F is the result of a combination of life table data and adjustment factors evaluated together. The method was, however, originally applied to well-documented plant species, which is not the case for the plant species of Maputaland. Because we believe that the method should be available to non-specialists and practitioners as well as being applicable to species for which only baseline survey data are available, we developed an empirical method to derive F. We first established the value of F (Step 1a), which is the minimum viable population size independent of the local conditions of the targeted region. In Step 1b, F is adjusted to reflect local conditions.
To determine F we used published data from detailed studies (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) of F-values against life expectancies of Banksia cuneata (Australia: F = 6,400, life expectancy = 50 years), Banksia goodii (Australia: F = 300, life expectancy = 300 years), Alnus incana (Northern Europe: F = 750, life expectancy = 20 years) and Pentaclethra macroloba (Amazon basin: F = 2,300, life expectancy = 100 years). Four equations were fitted to the available data to determine which best described the relationship: y = a(x) + b, y = a ln(x) + b, y = a(x)b, and y = aeb (x), where a and b are constants. The regression with the best fit (i.e. the highest coefficient of determination, R 2) was selected to calculate F for each of the landscape species selected for our case study.
Adjusted F is based on available knowledge regarding the species and environmental factors in the target area (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001). Thus, based on all available knowledge and expert opinion, F can be increased or reduced to accommodate the effects of environmental conditions in the target area on the species’ biology. We consider the value of F estimated in Step 1 to be the minimum viable population size and therefore adjusted F could only be ≥ F. Our decision to set the maximum adjustment value to 2F was subjective and based on personal knowledge of the area. Such subjective decisions are allowed in the original methodology as long as the person making the decision is reasonably qualified as an expert for the area concerned.
To establish the adjusted F-value we used the list of 25 ecological factors of Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001). Each factor had two alternative states: one related to species resilience and one to species vulnerability (Table 1). Each factor was investigated and a numerical value assigned per factor to derive an ecological factor score. This score was used to generate the adjusted F-value. If the available knowledge of the species was insufficient to obtain a reliable answer a value of 0 was given to both resilience and vulnerability. If it was possible to answer a question reliably then +1 was given for resilience or -1 for vulnerability. The ecological factor score was obtained by summing the resilience and vulnerability scores, and plotting the sum on a scale ranging from 25 (all questions pertaining to resilience assigned +1), thereby requiring no adjustment of F, to a low of -25 (all questions pertaining to vulnerability assigned -1) requiring maximum adjustment (F is doubled). The adjustment factor value was calculated as a fraction of the total potential for adjustment, based on the ecological factor score, and was added to F to obtain the adjusted F-value (Table 1).
Factors considered in the determination of the ecological factor score and thus the adjustment factor for the minimum viable population size (F), as required for Step 1 (Fig. 1) for each of the four selected landscape tree species (see text for details) in the Tembe Elephant Park–Tshanini Conservation Area complex, Maputaland (Fig. 2).
1 Represents any large scale, landscape shaping disturbance known to have occurred prior to the colonization of South Africa by Europeans
2 To calculate the adjustment factor the following scale is used: if EFS = +25 then AF = +0% of F, if EFS = 0 then AF= +50% of F, if EFS = -25 then AF = +100% of F. For example, for N. hidebrandtii the minimum viable population size (F) = 186.0, the AF value (40% of 186.0) = 74.4 individuals, and the Adjusted F value = 186.0 + 74.4 = 260.4. As we only work with whole individuals a rounded Adjusted F of 260.0 is used.
Step 2 The 36 × 42 km area encompasses both Tembe Elephant Park and Tshanini Community Conservation Area (Fig. 2). The region was not fully surveyed and therefore only available geographical elements known from field surveys are provided as mapped elements in Fig. 2. The unmapped sections were evaluated by analysing topographical maps and 1 : 50,000 aerial photographs. The disturbance level in each of the 42 squares was categorized as either sustainable (insignificant disturbance), light (light to moderate disturbance linked to human activity such as resource gathering) or heavy (moderate to heavy disturbance associated with human settlements and roads).
Tembe Elephant Park and Tshanini Community Conservation Area in Maputaland, the region evaluated for establishing a minimum viable area for conservation using the method of Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001). Individual grid squares are referred to in the text using the letters and numbers (e.g. D3). The area enclosed by the thick blue line indicates where development is expected to occur and to modify the landscape irreversibly within the next 50 years. The inset in the bottom right hand corner indicates the location of Maputaland in KwaZulu-Natal province, South Africa.
Step 3 The potential habitat per disturbance region (defined as the area capable of supporting viable populations of the selected landscape species, and characterized by climatic and/or environmental parameters necessary to support the successful establishment of the selected plant species) was evaluated either directly using knowledge from fieldwork within the surveyed areas or estimated when fieldwork-based knowledge was unavailable. Overall it was estimated that 29,000 ha could potentially support Sand Forest vegetation and 62,991 ha could support Woodland vegetation.
Step 4 All the potential habitat area that was physically surveyed was mapped, and consisted of the sum of the area of Tembe Elephant Park and of the disturbance region in which the Tshanini Community Conservation Area occurs. The total potential habitat surveyed amounted to 33,600 ha, of which 5,600 ha and 28,000 ha were surveyed in Sand Forest and Woodland vegetation respectively (Gaugris, Reference Gaugris2008). Of the 28,000 ha of Woodland, 12,200 ha was of Closed Woodland and 15,800 ha of Open Woodland.
Step 5 The density of mature trees per ha (D) was obtained from detailed field surveys (Gaugris, Reference Gaugris2008) in the Tshanini Community Conservation Area (disturbance region B6, Fig. 2) and Tembe Elephant Park (disturbance regions B1–B5, C1–C5, D1–D4, E1–E3 and F1–F2; Fig. 2). To account for human utilization, tree density outside these disturbance regions was taken as the weighted mean of the values of both conservation areas, reduced by 10%. To calculate D (Step 5; Table 2) in the Sand Forest, mature C. schlechteri and N. hildebrandtii trees were considered as those having a diameter ≥ 127 mm at breast height. In the Woodland, mature S. birrea and H. ulmoides trees were considered as those having a diameter ≥ 95.5 mm and ≥ 16 mm at breast height respectively.
Baseline information required to establish the minimum conservation areas for the four vegetation types studied in Maputaland (Fig. 2). For each of the four landscape species the following information is provided: the theoretical life expectancy (based on a literature search), the minimum viable population size (F, the number of individuals), the adjusted F value calculated to compensate for known life history characteristics of the species and environmental characteristics of the area (Table 1), and the mean density (number of individuals per ha) of mature trees in the Tshanini Community Conservation Area, Tembe Elephant Park and unprotected areas.
Step 6 The determination of the minimum area required for conservation started by establishing the ‘target area based on background disturbance processes, also called raw target area for reserve creation’ (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001), which was calculated as Target Area (A 0) = Adjusted F/D. This target area, calculated per disturbance region, is an absolute minimum area required to support the target species assuming there is no threat. It is a preliminary value that does not take into consideration other known disturbances that can be measured and planned for.
Step 7 Sociological studies, maps and photographs were used to determine the additional small-scale disturbances from which species are expected to recover within a 50-year time frame but that will effectively reduce the potential habitat available. These disturbances were expressed as a percentage of a disturbance region that would become unavailable during the time frame considered, from which the percentage of remaining habitat (S) of the disturbance region is calculated. This represents an evaluation of potential habitat that will become temporarily disturbed but that is considered likely to become suitable again provided enough time is given or appropriate management actions are taken. We considered here all activities that modify the canopy (cutting of trees) and sometimes the topsoil but not the seed bank (non-mechanized slash-and-burn).
Step 8 The area expected to become irreversibly damaged within 50 years through human developments (ci) was similarly evaluated per disturbance region and expressed as a percentage, and percentage of remaining habitat calculated as (1–ci). This represents an evaluation of potential habitat that will become irreversibly lost and will not become suitable again. We considered here all activities that will permanently modify the soil structure (construction and mechanized agriculture) and the presence of a seed bank.
Step 9 The target area per disturbance region was further adjusted to compensate for expected density-reducing human-related activities within the next 50 years and was expressed as an estimated percentage of remaining habitat (ri). This represents an evaluation of the effect of activities that will affect the species’ density in the landscape. We considered here the following three activities: building construction (through wood use for construction and as fuel), grazing (removal of seedlings), and weather effects (droughts and floods). Once this step was completed a final estimate of the minimum target area for conservation per disturbance region was thus obtained.
Step 10 was not applied as no catastrophic disturbances are expected in the region. Based on a review of historical, geological and weather records, no volcanic eruption, earthquake, landslide, tsunami, hurricane wind or tornado are expected.
Step 11 was also not applicable to our case study because it was already incorporated within our approach in 42 disturbance regions, whereby Sand Forest and Woodlands targets were combined.
Step 12 The area of potential available habitat is divided by the final value obtained for the minimum target area for each of the 42 disturbance regions. This ratio gives managers a rapid estimate of the current availability of land for conservation.
Results
The exponential equation (y = aeb (x)) provided the best fit (R 2 = 0.43) between F-value and life expectancy and was therefore used to derive F-values for the selected species (Table 2). H. ulmoides had the highest F-value at 1,086 individuals and N. hildebrandtii the lowest at 186. However, H. ulmoides required only a 14% increase of F to derive the adjusted F-value, whereas N. hildebrandtii required a 40% increase to obtain the adjusted F-value (Table 1).
From the density (D) and adjusted F, a target area representing the minimum viable area (Step 6) was calculated for each of the Sand Forest and Woodland species. The size of this area was then further refined based on the consideration of a group of reversible and irreversible disturbances likely to affect the area (Steps 7–9). See Appendices 1–2 for the full analysis.
The sum of all minimum and mean areas required for conservation per disturbance region for the entire region, as well as for the Tembe Elephant Park and Tshanini Community Conservation Area separately, are presented per species per vegetation unit (Table 3). The Sand Forest species required the smallest areas, with C. schechteri requiring a mean of 76 ha per disturbance region (range 8–1,706 ha), whereas N. hildebrandtii required a mean of 81 ha (range 5–1,808 ha). The mean minimum areas required for the Woodland species were greater. For H. ulmoides there was an increase in the area required from the Closed Woodland (mean 194 ha, range 148–4,349 ha) to the Open Woodland (mean 544 ha, range 5–9,752 ha), whereas the area requirements for S. birrea were similar in the Closed (245 ha; range: 10–5,411 ha) and Open Woodland (250 ha; range 28–5,657 ha).
Summary of the minimum area required for conservation per vegetation type and woody plant species in the whole study area, Tembe Elephant Park, and Tshanini Community Conservation Area (Fig. 2). The sum in hectares for each species represents the sum of all minimum areas for all disturbance regions in the site evaluated and that need to be conserved to have a satisfactory conservation outcome. The mean area represents the mean minimum area required for conservation per disturbance region. For Tshanini Community Conservation Area the mean could not be calculated as the site was encompassed within one disturbance region.
From Step 12 it is evident that the availability of Sand Forest habitat far exceeds the minimum area required to conserve this vegetation type in a 50-year time span (Table 4). Only in disturbance region D5 (Fig. 2) is the ratio of available to required habitat < 1. The Woodland analysis shows that some disturbance regions do not currently have enough habitat available (ratio < 1) for their conservation (Table 4). Of particular importance are disturbance regions B5, C5, D5, E2, E3 and F2, within Tembe Elephant Park, where insufficient habitat is available for at least one vegetation type and landscape species.
The ratio of available to required habitat per species (Step 12 of Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) in each of the evaluated disturbance regions (Fig. 2). A ratio > 1 indicates that sufficient available habitat exists to match the conservation requirements, whereas a ratio < 1 indicates that insufficient habitat is available. Numbers in bold indicate disturbance regions where insufficient habitat is available for conservation of that landscape species.
Discussion
One of the central questions in conservation science is ‘How much is enough?’ (Poiani et al., Reference Poiani, Richter, Anderson and Richter2000; Svancara et al., Reference Svancara, Brannon, Scott, Groves, Noss and Pressey2005; Tear et al., Reference Tear, Kareiva, Angermeier, Comer, Czech and Kautz2005). No universally accepted method to obtain the answer has yet been developed. We based our investigation of the conservation adequacy of the Tembe–Tshanini complex using a simple population model (Burgman et al., Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) to set targets for conservation. Population models have the advantage that they can deal with processes that affect the persistence of the species (Nicholson et al., Reference Nicholson, Westphal, Frank, Rochester, Pressey, Lindenmayer and Possingham2006) and are therefore ideal for single-species investigations but not for multiple species planning (Moilanen & Cabeza, Reference Moilanen and Cabeza2002).
For both Sand Forest species, C. schlechteri and N. hildebrandtii, it appears that sufficient habitat is already conserved in the Tembe Elephant Park (3,400 ha) and informally conserved in the Tshanini Community Conservation Area (1,046 ha; Gaugris, Reference Gaugris2008) to secure the future of both species and therefore the Sand Forest vegetation. Even in most of the disturbance regions covering land that is not conserved the minimum area necessary for conserving the species is present. This implies that suitable environmental conditions still exist beyond the conservation areas and that there is a functional corridor between the two areas. Disturbance region D5 is the only one in which the area available for Sand Forest was less than the required minimum. Most of this disturbance region (95%) is occupied by the village of Sicabazini, where the community resettled when they were relocated upon creation of the Tembe Elephant Park in 1983 (Matthews et al., Reference Matthews, Van Wyk, Van Rooyen and Botha2001). To ensure the maintenance of ecological processes and overall conservation of the Sand Forest we recommend that a suite of Sand Forest patches of varying sizes be preserved within a network of reserves rather than in one single reserve (Possingham et al. Reference Possingham, Ball, Andelman, Ferson and Burgman2000; Van Rensburg et al., Reference Van Rensburg, Chown, Van Jaarsveld and McGeogh2000a,b; Gaugris & Van Rooyen, Reference Gaugris and Van Rooyen2008).
For the Woodland vegetation type the minimum area requirements for the two species investigated differed substantially. We recommend using the species that requires the largest area. For the Closed Woodland, therefore, at least 10,303 ha should be set aside (based on the calculations for S. birrea), whereas 22,843 ha of the Open Woodland should be set aside (based on the calculations for H. ulmoides). However, within the Tembe–Tshanini complex there is only 7,386 ha of Closed Woodland and 9,576 ha of Open Woodland and thus insufficient land is presently conserved to ensure the persistence of these vegetation types over the period considered. Outside these conservation areas the Closed and Open Woodlands are well represented. Due to the marginal subsistence agriculture value of the soils of the area (Eeley et al., Reference Eeley, Lawes and Reyers2001) and the high rate of emigration from this area to coastal towns, where tourism drives employment opportunities (Peteers, Reference Peteers2005), these woodlands may nevertheless continue to survive outside conserved areas.
Burgman et al. (Reference Burgman, Possingham, Lynch, Keith, McCarthy and Hopper2001) advocated the use of their method in regions where there is little available information on the species or vegetation dynamics, as is the case for Maputaland (Gaugris, Reference Gaugris2008). In this context, the method fulfilled its promise to estimate the minimum area of each vegetation type that should be conserved. The minimum areas determined appeared realistic and it was possible to determine whether the areas currently set aside for conservation are adequate.
The weakest aspect of the method is that used to establish F, the minimum viable population size, without specialist knowledge or extensive life table information. A lack of detailed information (Cabeza & Moilanen, Reference Cabeza and Moilanen2001; Justus et al., Reference Justus, Fuller and Sarkar2008) or lack of reliability about the data, leading to uncertainty (Moilanen et al. Reference Moilanen, Runge, Tyre, Carmel, Fegraus and Wintl2006a,b), are two of the problems with which conservation planners have to contend. Although the models fitted to the published data did not have high R 2-values, the method provided a framework to calculate F objectively. Additional information would improve the fit. The adjusted F was derived by considering a range of ecological factors that are important for the long-term conservation of the species. These factors could readily be assessed using personal and expert knowledge. However, our attempts to establish F and adjusted F are preliminary and require revision. Future studies should focus on devising a simple and repeatable method of setting F without knowledge of growth, mortality and recruitment rates. Perhaps a way of setting F from size class distribution and abundance data could be investigated.
Species selection is crucial because the species are used as surrogates for conservation of an entire vegetation type. A common conservation strategy has been to use umbrella species as surrogates for poorly known regional biota (Simberloff, Reference Simberloff1998), similar to the use of indicator taxa in reserve design (Howard et al., Reference Howard, Viskanic, Davenport, Kigenyi, Baltzer and Dickinson1998). However, evaluations of the use of umbrella or flagship species have indicated that they perform no better than randomly selected species (Andelman & Fagan, Reference Andelman and Fagan2000; Williams et al., Reference Williams, Burgess and Rahbek2000), although encouraging results have been obtained in Uganda (Howard et al., Reference Howard, Viskanic, Davenport, Kigenyi, Baltzer and Dickinson1998). The selection of landscape plant species appeared to work adequately in this study. We recommend, however, that future studies should select one species per vegetation unit, rather than trying to bridge more than one vegetation unit with one species.
A major advantage of the method is its ease of use once it has been established for an area. The wide variety of quantitative and qualitative aspects considered allows further refinements to be made as new knowledge becomes available. The method can be adapted to include additional aspects (within Steps 7–9) such as agriculture, alien plants, urbanization and interface retention targets (Pressey et al., Reference Pressey, Cowling and Rouget2003), and additional aspects of the population biology of the species can be included to refine the adjusted F-value. By changing the values in the tables (Appendices 1–2) instantaneous results are obtained and therefore decisions can be made without extensive consultation. Using the tables a conservation manager can establish the minimum area of land that should be fully preserved in each disturbance region, and thus establish the area remaining for sustainable use. An additional advantage of the method lies in the possibility to scrutinize an area using a range of disturbance region sizes and therefore to balance conservation according to the scale of definition desired. We feel that the advantages of such a tool for conservation, especially outside formally conserved areas, justify applying the method even if data are deficient initially, if only to establish targets for conservation and areas for sustainable use.
A clear limitation must, however, be set on the size of the area under scrutiny. It should remain within the grasp and knowledge of the ecologist and experts tasked to set the relevant minimum viable areas. Should the area become exceedingly large the conditions are likely to be beyond the knowledge of the experts. This method appears suitable for setting targets by local people and experts. However, its applicability at a larger scale is more uncertain as a lot of definition would be lost. It would be possible to investigate the applicability of the method for a larger scale based on the results of many subregions, each considered as disturbance regions for the larger scale meta-analysis.
To conclude, the method described here is promising as it sets applicable limits and is relatively easy to use, replicate, update and refine. The total area that we considered was approximately five times larger than the combined areas of the reserves concerned, and we feel that the local dynamics of the ecosystems were appropriately represented within this total area.
Acknowledgements
This article is dedicated to Irmie Gaugris, who left us too early. We thank Caroline Vasicek for her insight and for reviewing the methodology, and four anonymous referees for their valuable comments. This research was supported by the National Research Foundation (South Africa) under Grant Number 2047386.
Appendices 1–2
The appendices for this article are available online at http://journals.cambridge.org
Biographical sketches
Jerome Gaugris is an environmental consultant specializing in the sustainable utilization of natural resources in sub-Saharan Africa. He currently works on ecosystem service delivery in a highly populated urban landscape in Burundi. Gretel Van Rooyen specializes in plant adaptations to desert environments, with a particular focus on the Namaqualand region, and also supervises a conservation-based rural community development programme in Maputaland.
