Exploring the mesofilter as a novel operational scale in conservation planning


Correspondence author. E-mail: Casparus J. Crous, casperc@sun.ac.za


  1. Increased emphasis is being placed on developing effective biodiversity conservation tools for practical conservation planning. The mesofilter is such a biodiversity planning tool, but has yet to be fully explored to appreciate its effectiveness. The key premise of the mesofilter is that ecosystems contain certain physical elements that are specifically associated with a diversity of species. Identifying such mesofilters could therefore complement existing conservation planning tools such as coarse and fine filters.
  2. To explore the value of the mesofilter as an operational scale in conservation planning, we studied 18 remnant patches of endangered montane grassland in KwaZulu-Natal, South Africa, using the physical landscape feature of patch rockiness as an abiotic surrogate for biodiversity. The objective was to determine whether the mesofilter of rockiness can predict variation in species richness and composition for three dominant grassland taxa (plants, butterflies and grasshoppers) at the landscape scale.
  3. Variable levels of rockiness had significant interactions with all three focal taxa. Higher species richness of all taxa was closely associated with higher levels of rockiness in a patch. The rocky mesofilter only predicted significant differences in species composition for butterflies. Elevation was also important, possibly another mesofilter for plants and grasshoppers in this landscape.
  4. Synthesis and applications. The results indicate that the use of an abiotic surrogate such as rockiness can predict biodiversity value across multiple taxa. The mesofilter is therefore a valuable surrogacy and congruency tool for practical biodiversity conservation across this landscape and would likely have similar value if explored elsewhere. It also has value in the design and management of protected areas.


One of the main goals of systematic conservation planning is to encapsulate the complexity of biodiversity across different spatial scales and geographical regions when delineating a protected area (Margules & Pressey 2000; Pressey et al. 2007). To address this complexity, many biodiversity conservation tools have been developed. These focus on designating a protected area using different species and habitat heterogeneity concepts (Schulte et al. 2006). Of these, fine- and coarse-filter operational scales are often used to delineate networks of protected areas (Noss 1987; Schwartz 1999). Protected areas are either designated for a specific species, usually a flagship one, or around a set geographical area, for example, 1000 km2 of a certain ecosystem (Noss 1987). However, both these fine- and coarse-filter operational scales have their shortcomings.

Fine-filter approaches usually entail the use of surrogates of biodiversity through concepts such as umbrella species, focal species or even guilds (Marcot & Flather 2007). However, congruency issues arise when these surrogates do not adequately represent targeted taxa or overall biodiversity (van Jaarsveld et al. 1998; Lindenmayer et al. 2002). This means that using focal species as a proxy to protect other taxa could be problematic, as species-specific requirements towards habitat conditions, and their response towards threats, are highly variable in space and time (Lindenmayer et al. 2002). Also, areas that are poorly surveyed might lead to false absence of a species and consequently be mistakenly excluded from protected areas (Ferrier 2002). Therefore, in many circumstances, fine-filter conservation is not the appropriate approach, because what is needed is to select surrogates (and subsequently protected areas) in such a way that it will also ensure that spatial autecological requirements of most species are met (Margules & Pressey 2000).

In contrast, coarse-filter reserve selection is theoretically more directed towards including multiple ecosystem types or cover types. However, the problem with coarse-filter approaches is that in most cases, a lack of knowledge may lead to protected areas not being truly representative of natural ecosystems (Margules, Nichous & Pressey 1988) and, in doing so, fail systematic conservation planning. Therefore, for many protected areas to persist, they often need to be expanded into the surrounding matrix to encompass these spatial autecological deficiencies. This can be problematic due to ongoing human infrastructure development (Maiorano, Falcucci & Boitani 2008).

To address this disparity in conservation planning, Hunter (2005) developed a new operational scale for biodiversity conservation – the mesofilter. Broadly, the mesofilter can be defined as specified ecosystem elements, or features, which are important for the maintenance of certain species within an area. The mesofilter complements the coarse filter by helping conservation planners to delineate those physical features of the landscape that are known to be associated with, and promote, a higher diversity of species (Hunter, Jacobsen & Webb 1988). Furthermore, the conservation significance of using this complementary approach to conservation planning is highlighted, as many mesofilters could also endure over long periods, despite climate change (Hunter, Jacobsen & Webb 1988). Therefore, this mesofilter approach at least partly overcomes the flaw in fine-filter conservation, by focusing on those ecosystem elements that are easier to survey and map than single species. Conversely, instead of using biotic components as surrogates for other biota, the emphasis here is on the use of abiotic elements as surrogates for biota (Carroll 1998). The mesofilter ensures that protected area selection, as well as selecting conservancies outside protected areas, incorporates multiple environmental elements within the geographical area to ensure more comprehensive conservation of biodiversity, compared with an area adjacent or nearby which lacks these elements.

However, the mesofilter concept has not to date received much attention as an operational scale in conservation planning. Many studies have shown certain habitat elements or landscape features to be important indicators of diversity, emphasizing that conservation of these elements leads to protection of a diversity of species (Armstrong, van Hensbergen & Geertsema 1994; Armstrong & van Hensbergen 1999; Wessels, Freitag & van Jaarsveld 1999; Hewitt et al. 2005; Overton, Schmitz & Casazza 2006; Barton et al. 2009; Overton, Casazza & Coates 2010). Barton et al. (2009), for example, showed that woody logs in a reserve area had specific associations with many beetle species. These logs increased the biodiversity of the area, so delineating beetle biodiversity hotspots. This is important for protected area design and management, as incorporating these logs as part of the conservation planning will increase biodiversity at the landscape level. Therefore, the mesofilter provides a practical approach to inventorying landscape features of increased biodiversity value, to which subsequent management could be directed (Lindenmayer et al. 2008). Similarly, should a new protected area network be designed, identifying habitat elements that provide a characteristic assemblage of species would prove a vital addition to the design of the conservation network. The efficacy of using a similar complementary approach when designating biodiversity hotspots within a protected area has been shown (Noss et al. 2002). Recognizing mesofilter conservation per se, as posited by Hunter (2005), therefore needs to be explored, particularly as it shows promise as a valuable new operational scale in the biodiversity and conservation planning toolbox (Schulte et al. 2006; Samways, McGeoch & New 2010).

In South African montane grasslands, Armstrong, van Hensbergen & Geertsema (1994) provided some evidence that rockier landscapes had higher plant and butterfly species richness. Here, we assess the value of mesofilters for conservation planning by looking at this rocky mesofilter. To achieve this, we explore whether percentage rockiness in this case (juxtaposed to elevation as a proxy for microclimatic variation) can predict patterns of varying plant, butterfly and grasshopper species richness at the landscape scale, and in addition to species richness, determine the influence of these habitat characteristics on the similarity of species assemblages across this landscape.

Materials and methods

Study area

The study was conducted within the 16 000 ha Merensky Forestry Estate at Weza, near Kokstad, KwaZulu-Natal, South Africa (30°34·855 S, 029°44·726 E; Fig. 1). Around 4 200 ha are semi-natural open spaces, the remainder being commercial forestry. The open spaces lie mostly within the endangered Midlands Mistbelt Grassland vegetation type (Mucina & Rutherford 2006). The endangered status of this vegetation type is mainly driven by large forestry plantations and activities in the area. The dominant grass in the area is Themeda triandra Forssk. All selected sites are classified as semi-natural, as all were annually burned by forestry management over six decades. Moreover, grazing is limited within these remnants, and fire is consequently seen as the main ‘herbivore’ (Bond & Keeley 2005). To avoid pseudoreplication, sites of higher rockiness were interspersed with those of lower rockiness across the study area, with the minimum distance between similar sites being 400 m. In addition, all sampling occurred >30 m away from the pine forest edge, to reduce sampling bias due to edge effects (Samways & Moore 1991; Bieringer & Zulka 2003; Pryke & Samways 2012).

Figure 1.

Location of the Merensky Forestry Estate at Weza, KwaZulu-Natal Province, South Africa. Indicated numerically are the sampling sites, all within the open semi-natural grassland areas.

Flora sampling

Eighteen flora sampling sites were selected. Sampling was carried out between January and February 2011 (Armstrong, van Hensbergen & Geertsema 1994), through a fixed-grid sampling design, where sampling is taken at fixed intervals along a determined gradient (Whalley & Hardy 2000). This design is relatively easy to perform in the field and has been shown to obtain data rapidly on species distribution and abundance within a study area (Tucker et al. 2005). Within this design, we used point intercept line transects, as this method has been shown as relevant and insightful for biodiversity studies in these grasslands (Everson & Clarke 1987; Armstrong, van Hensbergen & Geertsema 1994).

Field methods were similar to Hayes & Holl (2003), where a 50-m measuring tape was used to record all plant species that intercept a 1·8-mm diameter pin every 1 m (51 points per transect). For grasslands, a dense vegetation type, transects of 50 m are seen as adequate (Rich et al. 2005). A total of four 50-m transects were placed within each of the eighteen sites, each transect being 15 m away from another, effectively having 204 points per site. Percentage rockiness was obtained by adding the number of times a rock (any rocky surface >10 cm in diameter) touched the pin divided by the total number of pin hits per transect. Also, a metal stake was inserted in the ground every 5 m on each transect, giving 40 measurements per site, which serves as a composite indicator of surface rockiness (Stohlgren & Bachand 1997). We then correlated the soil depth with percentage rockiness to ensure correct classification of the site as rocky, and not just a rocky outcrop within a non-rocky matrix.

In addition, a 1-m belt, perpendicular to the line transect, was time-searched for 15 min after each transect measurement, as a means for recording a more comprehensive species list that could include short-lived annual plants (Hayes & Holl 2003).

Butterfly sampling

Butterfly sampling was at the same 18 sites as the flora sampling. Butterflies were sampled twice, in January and April 2011, to encompass seasonal differences. They were sampled within a 50-m radius from the middle point of each site, by two observers facing opposite directions. Each observation unit was 30 min and replicated over three different days, at three different times of the day, making 90-min search time per person per site (3 h total per site). Sampling was between 09·00 and 15·00, on warm or hot days (average temperature of 30·2 °C for January counts, and 24·7 °C for April counts) with <5% cloud cover. To obtain butterfly species richness per site, observations from all replicates were pooled.

Grasshopper sampling

Grasshopper sampling was at the same 18 sites as the flora and butterfly sampling. Sampling was conducted twice, January and April 2011, to cover seasonal differences. Grasshoppers were sampled by sweep netting, which is adequate for short dense vegetation types such as grasslands (Gardiner, Hill & Chesmore 2005). Two 100-m transects were laid out. Parallel to each side of each transect, a 100 180° sweeps were made with a mesh net (diameter 40 cm). This rendered 200 sweeps per transect and ultimately 800 sweeps per site across the study period.

Statistical analysis

To ensure adequate taxon representation, sampling was conducted until the species accumulation curve nearly flattened (Gotelli & Colwell 2001). Data were then divided into two sets: continuous data for regression analysis and generalized linear modelling, and categorical data for analysis of variance (anova) and permutational multivariate analysis of variance (permanova) statistics. Pertaining to categorical data, both the rockiness and elevation values were tested for normality and their variances tested for homogeneity using a Shapiro–Wilk test (statistica Release 10; StatSoft, Inc., Tulsa, OK, USA). In both instances, the points were normally distributed around the means. As such, there were no distinct groups, and percentage rockiness was presented as a binary classification based on areas having more or <10% rockiness, as this was close to the average percentage rockiness measured across the 18 study sites (data not shown). Similarly, elevation was presented as a binary classification established at higher or lower than 1280 m a.s.l., as this was the average measured elevation across the 18 study sites (data not shown). The data were also categorized in this instance to have a practical example of possible implementation in the field.

To examine the overall relationships between richness of all recorded species per site and the measured environmental variables, scatter plots reporting r-values were constructed (statistica Release 10; StatSoft, Inc.). Similarly, to observe the relationship between each taxon and the measured environmental variables, scatter plots reporting r-values were constructed. To further explore the contribution of the environmental variables on species richness and abundance, we made use of generalized linear models (GLZ) (McCulloch, Searle & Neuhaus 2008) in statistica Release 10 (StatSoft, Inc.). For flora and grasshopper species richness, each GLZ had a normal distribution and an identity-link function. For butterfly species richness, a Poisson distribution with a log-link function was used. For abundance data, all tests were carried out with a Poisson distribution and a log-link function.

To examine the possible combination of factors driving differences in species richness in space, the data set was then divided into four groups with regard to habitat rockiness and elevation. These groups were as follows: high (elevations > 1280 m a.s.l.) with >10 (areas with more than 10% habitat rockiness), high (elevations > 1280 m a.s.l.) with <10 (areas with <10% habitat rockiness), low (areas <1280 m a.s.l.) with >10 (areas with more than 10% habitat rockiness) and low (areas <1280 m a.s.l.) with <10 (areas with <10% habitat rockiness). Species richness for all measured taxa across these groups was compared statistically using a factorial anova followed by a Fisher LSD post hoc test (statistica Release 10; StatSoft) to identify any between-group differences. Data were transformed where necessary to adhere to statistical models.

Finally, to explore whether differences in species composition across study sites (if any) could be a function of habitat rockiness or elevation, we used canoco 4.5 (ter Braak & Šmilauer 2002) and permanova (Anderson 2001) in primer 6 (PRIMER-E 2008). In canoco, we made use of canonical correspondence analysis (CCA) to explore the overall effect of percentage rockiness and elevation on taxa assemblage composition. We also overlaid species richness as a descriptive supplementary variable on each CCA. Forward selection during the CCA analysis was used to rank the most important environmental variables that structure species distribution within each taxon. We used permanova to study whether there were differences in species assemblage composition across our experimental rockiness and elevation categories. For this statistical method, we used an overall test, comparing species composition across each factor (rockiness and elevation), and pairwise tests (comparing species composition within different levels of both factors combined, with categories parallel to the ones used for the species richness anova test). permanova results are reported as P-values (e.g. McNatty, Abbott & Lester 2009), where a significant P-value indicates a significant difference between two levels (groups) of a studied factor. Analyses were performed using Bray–Curtis similarity measures where data for each taxon was fourth-root-transformed to reduce the weight of the common species (Anderson 2001).


Species richness and abundance across environmental variables

A total of 317 plant species (6574 individuals), 47 butterfly species (551 individuals) and 48 grasshopper species (864 adult individuals) were sampled. Overall, percentage rockiness showed a strong positive correlation with total species richness per site (three taxa combined) (r = 0·84, < 0·001), whereas elevation showed no significant correlation (r = −0·38, = 0·12) (Fig. 2). Percentage rockiness also had no relationship with elevation (r = −0·08, = 0·76) (Fig. 2). More specifically, percentage rockiness explained a significant part of the variance observed in both flora (r = 0·806, < 0·05) and butterfly (r = 791, < 0·05) species richness across the study sites (Fig. 3). Elevation had a statistically significant relationship only with grasshoppers (r = −0·514, < 0·05) (Fig. 3).

Figure 2.

The relationships between % rockiness, elevation and the total number of plant, butterfly and grasshopper species recorded at each site. = 18.

Figure 3.

The relationships between plant, butterfly and grasshopper species richness and elevation and % rockiness in a patch. = 18.

Furthermore, results from the generalized linear modelling (GLZ) showed the significant influence of both percentage rockiness and elevation on the species richness of flora and grasshoppers (Table 1). However, for flora, percentage rockiness had a stronger effect than elevation, whereas for grasshoppers the opposite was true. In contrast, percentage rockiness was the only variable that significantly influenced butterfly species richness (Table 1). Grasshopper abundance was significantly influenced by both elevation and percentage rockiness (Table 1). As with species richness, butterfly abundance was only significantly influenced by percentage rockiness (Table 1). None of the two tested variables significantly influenced floral abundance.

Table 1. Generalized linear modelling (GLZ) for species richness and abundance of taxa, showing their relationship with measured environmental variables
TaxonVariabled.f.Wald statisticP-value
  1. Values in bold are significant at the 5% level.

Species richness
FloraElevation16·70 0·010
% Rockiness142·74 <0·001
% Rockiness110·81 0·001
GrasshoppersElevation17·37 0·007
% Rockiness15·42 0·020
% Rockiness12·620·106
% Rockiness169·78 <0·001
GrasshoppersElevation146·24 <0·001
% Rockiness129·90 <0·001

For flora, mean species richness differed significantly between categories (Fig. 4a) and was mainly driven by the significant decrease in species richness observed for areas that had <10% rockiness. In particular, the category ‘high elevation with <10% rockiness (High < 10)’ had on average lower species richness than all other categories and significantly lower species richness than both areas of higher percentage rockiness. This result for flora was the same for grasshoppers (Fig. 4b). In contrast, butterfly species richness did not differ significantly across any of the categories (Fig. 4b). However, butterfly species richness was on average the highest in areas with higher percentage rockiness.

Figure 4.

Mean (±SE) for (a) flora and (b) butterflies (light grey) and grasshoppers (dark grey) relative to measured environmental variables. High represents sites >1280 m a.s.l. and low <1280 m a.s.l. >10 represents areas that are >10% rocky and <10 areas lower than 10% rocky. Within taxa, means with different alphabetical letters differ significantly (< 0·05).

Species composition relative to measured environmental variables

Canonical correspondence analyses (CCA) revealed that assemblages of both flora and grasshoppers were more strongly structured in space by elevation than by percentage rockiness (= 0·004 and = 0·287, respectively) (Fig. 5a,c). In contrast, butterfly assemblage composition was more strongly influenced by percentage rockiness (= 0·089) as opposed to elevation (= 0·256) (Fig. 5b).

Figure 5.

Canonical correspondence analysis for (a) flora, (b) butterflies and (c) grasshoppers across sites, the two measured environmental variables and a descriptive supplementary variable. Forward selection results showed that for flora, elevation had was a more significant influence than% rockiness (elevation, = 0·004;% rock, = 0·287); for butterflies,% rockiness was a stronger influence than elevation (elevation, = 0·256;% rock, = 0·089); and for grasshoppers, elevation was more significant than percentage rockiness (elevation, = 0·001;% rock, = 0·881). FSR, floral species richness; BSR, butterfly species richness; GSR, grasshopper species richness.

Similar to the canoco results, but with using our experimental categories, the only significant interaction between percentage rockiness and focal taxa composition was for butterflies (permanova,= 0·002; Table 2). In turn, flora and grasshoppers were the only taxa that showed significant differences in assemblages relative to elevation (permanova,= <0·001 and = 0·010, respectively; Table 2). Pairwise tests showed that for flora, the combined group of high elevation sites with <10% rockiness was consistently driving the differences in species composition across sites (Table 2). Similar results were obtained for grasshoppers, although this result was not as pronounced as that of flora. In contrast, the butterfly assemblage was not at all influenced by this combination of environmental variables. Instead, they were more strongly influenced by lower elevation areas with <10% rockiness (Table 2).

Table 2. Permutational multivariate analysis of variance (permanova) results on the effect of elevation and percentage rockiness per habitat on species composition for three taxa
  1. High represents sites >1280 m a.s.l. and low < 1280 m a.s.l. >10 represents areas that were >10% rocky and <10 areas <10% rocky. Values in bold are significant at the 5% level.

Overall test P-valueP-valueP-value
Rockiness0·0532 0·0024 0·1318
Elevation <0·001 0·2822 0·0101
Rockiness × Elevation0·13590·82010·3157
Pairwise test
>10 high, <10 high 0·008 0·32530·1089
>10 high, <10 low0·5612 0·0073 0·2533
>10 high, >10 low0·82570·85540·1715
<10 high, <10 low 0·0084 0·295 0·0338
<10 high, >10 low 0·0068 0·2922 0·0168
<10 low, >10 low0·8099 0·0082 0·4635


Since the inception of the mesofilter concept (Hunter 2005), little research has been carried out to explore this as a practical field tool. Moreover, little research has been carried out to explore the relationship between physical ecosystem features and species richness and composition for practical conservation planning. Here, we tested the use of rockiness as a mesofilter as described by Hunter (2005). This physical landscape feature had significant interactions in species richness and composition with all three focal taxa. This interaction illustrates how we can apply environmental data using a mesofilter to help optimize design of conservation plans, and thus management of biodiversity, at a landscape scale.

Overall, the percentage of rockiness is an important driver of the variation observed here for species richness across all studied taxa. This result was true whether using either continuous data or our experimental categories. In fact, using a specific delineation of higher or lower than 10% rockiness, it was sites with <10% rockiness where all taxa had lower species richness. Moreover, it is clear that plant and grasshopper species richness were also influenced by elevation. We could thus infer that rockiness and elevation are important variables in delineating biodiversity hotspots for plant and grasshoppers. However, species richness alone is a poor indicator of biodiversity as a whole (Purvis & Hector 2000). For example, if one area had ten species and another had twenty, by using only species richness one might argue that the area with twenty species is more important to conservation. However, if the ten species found in the other site were significantly dissimilar in composition to the other area, both areas are indeed important for biodiversity conservation. In that sense, we also measured similarity/dissimilarity in species composition and whether this difference could be a function of the rockiness and/or elevation mesofilters, and what this would mean for biodiversity planning. Both flora and grasshoppers showed significantly different species composition for elevation, but not rockiness. Specifically, for both taxa, it was the combination of >1280 m a.s.l. and <10% habitat rockiness that influenced this observed assemblage difference. Essentially, for flora and grasshoppers at a landscape scale, we can more readily predict biodiversity ‘microhotspots’ (Grant & Samways 2011) as the rockiness mesofilter was strong enough to delineate biodiversity hotspots across two taxa. Furthermore, these results also emphasize the significance of both rockiness and elevation as mesofilters for delineating areas of conservation concern for plants and grasshoppers within this montane landscape.

The real question underlying these results is why these taxa would respond to the rocky mesofilter. Within grasslands, variable levels of fire disturbance are known to structure plant communities (Bond & Keeley 2005). Rocks within a landscape are implicated in lessening the severity of fires and are thus creating refugia for certain fire-sensitive species (Signell & Abrams 2006). Furthermore, rocks may also provide barriers against ground-dwelling herbivores that eat bulbous plants, again promoting the longevity of certain plants in rocky landscapes (Thomson et al. 1996). In grasshoppers, rocks are important structures that aid in thermoregulatory processes (Chappell 1983). Essentially, processes such as fire, predation and thermoregulation, which occur across many ecosystems, could all be seen as confounding variables in the response of taxa to rockiness. However, as this was not explicitly tested here, it remains to be fully explored for this grassland landscape.

Interestingly, in this montane landscape, elevation had no significant influence on butterfly species richness. Grill et al. (2005) also found moderate elevation differences to have no relationship with butterfly species richness and suggested that increased butterfly richness is in response to variation in favourable floral composition and structure. In contrast, variation in butterfly species richness has been shown to be a function of elevation and topographical heterogeneity (Mac Nally et al. 2003; Gutiérrez Illán, Gutiérrez & Wilson 2010). Overall, it seems that the factors influencing butterfly species richness might be a complex interaction between land cover heterogeneity, climate and topography (Kerr, Southwood & Cihlar 2001; Gutiérrez Illán, Gutiérrez & Wilson 2010). This means that diversity in land cover (measured at the large spatial scale) influences species composition in space, owing to different species inherently being associated with specific conditions (Fleishman et al. 2001). This would then ultimately explain the variation in species richness when measured at a small scale (Kerr, Southwood & Cihlar 2001). Consequently, butterfly species richness is a weak measure for delineating biodiversity hotspots at a small spatial scale, owing to high species turnover across a heterogeneous landscape.

The result from the butterfly permanova analysis supports the view that species richness alone is not an accurate indicator of biodiversity as a whole when measured at a small spatial scale. Percentage rockiness showed a strong influence in structuring dissimilar butterfly assemblages across this space. Thus, butterfly biodiversity microhotspots could not be predicted using the mesofilter. Nevertheless, this approach predicted whether a certain butterfly species is present or not. In other words, a certain assemblage of butterflies would be strongly associated with rocks, while another assemblage would be absent from such areas. The reason for this behaviour in butterflies remains to be fully explored. Still, this result has important conservation planning implications at the spatial scale of the landscape, as changes in species composition for butterflies are strongly influenced by rockiness (see Hewitt et al. 2005). This then enables a planning approach where certain landscape features and characteristics, as preferred by different taxa, could be incorporated into the systematic conservation planning process (Margules & Pressey 2000; Lindenmayer et al. 2008).

Subsequently, the biotic surrogacy issues, as raised by Lindenmayer et al. (2002) and Ferrier (2002), could also be addressed through using this rocky mesofilter. Here, the focus was on using abiotic surrogates. Lindenmayer et al. (2002) argued the probable failure of a focal species approach towards surrogacy, as habitat conditions are mostly variable, and therefore, species-specific requirements may also vary. Here, we kept the focal mesofilter constant. When more than one taxon is significantly associated with this mesofilter across space, whether through species richness or composition, as we show here, conservation planners can be more precise in knowing that species-specific requirements are kept constant across an area.

A further point is the importance of developing conservation planning tools, such as surrogates, which are likely to persist across different management regimes or environmental conditions (Hunter, Jacobsen & Webb 1988; Sarkar et al. 2006). In other words, surrogates need to be robust and designed so that they are consistently associated with their target species or taxa irrespective of habitat conditions due to varied management (e.g. between protected areas and unprotected remnant patches). Armstrong, van Hensbergen & Geertsema (1994), studying natural habitats, showed that plant species richness within montane grasslands in South Africa was higher in rocky areas. We have also provided significant evidence for this also being the case in semi-natural montane grassland remnants. In essence, the mesofilter concept, as proposed through this rockiness proxy, fits this recommendation for more accurate surrogates (Sarkar et al. 2006). Moreover, rocks are physical ecosystem features that persist over long periods, despite climate change, again emphasizing the mesofilter concept as a novel complementary approach to modern conservation planning (sensu Hunter, Jacobsen & Webb 1988). This highlights the value of a rockiness mesofilter as a conservation tool for this critically endangered habitat type in South Africa and is likely to have a similar value if explored elsewhere.

An important question remaining is whether abiotic factors are generally important to conservation. The mesofilters of rockiness and elevation studied here suggest that it is, but are not the ‘be all and end all’ for conservation planning, as many other features might also exist within a landscape, which would be as valuable to take into account. For example, different soil types were shown to be an important abiotic variable to take into account for conservation planning in prairie ecosystems in the United States (Wilsey, Martin & Polley 2005). Similarly, logs in Yellow Box–Red Gum grassy woodlands in Australia were shown to have high beetle diversity, which was particularly important towards conservation planning for this taxon (Barton et al. 2009). The importance of abiotic variables in an aquatic environment has also been reported, where piles of shell debris can significantly enhance diversity (Hewitt et al. 2005). Soil type, logs and shell debris are therefore mesofilters within their respective landscapes. Essentially, any ecosystem can be thought of theoretically having many attributes or features that would be of conservation interest, and mesofilters are therefore a way of expressing this attribute to be used in wildlife conservation evaluation (Usher 1986). A particular mesofilter we delineate is therefore an important departure point from which we start conservation planning within a landscape in a rapidly changing environment.


There is an increasing need to understand the determinants of observed spatial heterogeneity in species richness and composition (whether at a large or small spatial scales), as this will greatly optimize conservation planning for both biodiversity maintenance and the movement of species under a changing climate (Gaston 2000). This study presents a mesofilter approach, which adds to our current understanding of species distribution pertaining to certain landscape elements across a small spatial scale. Ultimately, the novelty arose by using an abiotic indicator approach, based on landscape elements that are easy to quantify and map and which are associated with multiple taxa. This would ease land-use decision making in similar areas where species inventories are currently lacking and development is taking place rapidly (Carroll 1998; Fleishman et al. 2001; Mac Nally et al. 2003). We strongly argue the value and relevance of this mesofilter operational scale to be used alongside the currently implemented conservation planning operational scales such as fine- and coarse-filter approaches (sensu Hunter 2005; Schulte et al. 2006).


We thank Merensky Forestry at Weza for access to the estate and Ezemvelo KZN Wildlife for a permit (no. 342/2011). Also we would like to thank J. Groenewald and L. van der Mescht for field assistance and C. Bazelet for help with grasshopper identification. CJC was supported through a grant from the Hans Merensky Foundation. Partial financial support to MJS was from the National Research Foundation (South Africa). We also thank the reviewers for their constructive criticism, which greatly improved the manuscript.