African Journal ofHerpetology. 2009 58(2): 71-84. ©HerpetoJogical Association ofAfrica

Original article The impacts of past cultivation on the in a South African grassland

GAVIN P. R. MASTERSON, BRYAN MARITZ, DARIAN MACKAY & GRAHAM 1. ALEXANDER

School ofAnimal, Plant and Environmental Sciences, University ofthe Witwatersrand, Johannesburg, . Email: [email protected]

Abstract.- Habitat transfonnation is the primary anthropogenic threat to global biodiversity. Fragmentation ofreptile populations following habitat transfonnation within a landscape can lead to the extirpation of species. We investigated the effects of land-use on the species richness and abundance of assemblages in three habitat types (two natural and one modified) in the grasslands of Gauteng, South Africa. Using trap arrays, we surveyed reptiles in primary grassland with little or no rock cover, primary grassland with large quartzite outcrops and scattered rocks, and secondary grasslands that were historically ploughed and cropped. We measured vegetation height and vegetation cover at these same localities. We caught significantly fewer reptile species in the historically cultivated sites than in either of the two natural habitat types. Differences in the reptile assemblage of each habitat type were not explained by either the spatial location or the vegetation structure of our trap sites but were well explained by the sites' habitat type. Estimates of total species richness indicated that we were able to adequately sample the reptile assemblages in the three habitat types, further supporting our observation ofreduced species richness in the secondary grasslands. We infer that habitat transfonnation associated with cultivation e.g., rock removal, has had a detectable, negative impact on the species richness and composition ofthe local reptile assemblages. We recommend that land-use planning in Gauteng empha- sise the need for areas ofinter-connected, untransfonned habitat in order to mitigate the negative impacts ofhabitat transfonnation on the local reptile diversity.

Key words.-Reptile, species richness, grassland, cultivation, sample-based rarefaction.

he loss or degradation of natural habitat cial in areas where human activities are dense- Tand ecosystems is the single biggest cause ly concentrated, e.g. Gauteng, South Africa. of biodiversity loss in South Africa's terrestrial ecosystems (Driver et al. 2005). Driver et al. Gauteng occupies only 1.4 % (17 0I0 \an2) of (2005) quantified the spatial location and South Africa's land area but is home to 21.5 % severity of threats to the country's biodiversity (l0.4 million) of its estimated 48 million peo- but did not assess the potential impacts ofland- ple (Statistics South Africa 2008) and produces use change on each ofthe components of bio- 33.9 % of South Africa's Gross Domestic Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 diversity. The impacts of land-use change and Product (GDP; Department of Environmental habitat modification/degradation on the reptiles Affairs and Tourism 2005). Between 2001 and of South Africa are currently poorly described 2008, the population density of the province (Smart et at. 2005; Masterson et al. 2008), rose from 520 people per \an2 (Department of despite evidence suggesting that reptiles are Environmental Affairs and Tourism 2005) to particularly sensitive to land-use (e.g. 616 people per \an2 (Statistics South Africa Santelmann et al. 2006). Research into the 2008). This rapid human population growth is impacts of land-use change is particularly cru- placing increasing pressure on the natural envi-

71 AFRICAN JOURNAL OF HERPETOLOGY 58(2) 2009

ronment in Gauteng. In addition to the land (around 60 %; Driver et at. 2005) ofall the nine allocated to meet the housing and infrastructur- recognised biomes in South Africa. Several al needs of the population, large areas of the grassland-specialist reptile species are declin- province's sweet grasslands have been cleared ing in abundance and Area of Occupancy e.g., to produce maize and other agricultural prod- Chamaesaura anguina anguina and ucts (Department ofEnvironmental Affairs and Chamaesaura aenea as well as Homoroselaps Tourism 2005). The present rate of habitat dorsalis (see M. Bates, unpub!.; Branch 1988). transformation shows no sign of declining in To date, the presence of the two Chamaesaura the short or medium-term, which emphasises species has not been confirmed in any ofthe six the need for conservation authorities to identi- provincially protected areas of Gauteng, while fY the impacts ofcontinuing transformation on H. dorsalis is known only from Suikerbosrand biodiversity and develop plans to mitigate Nature Reserve (Whittington-Jones et a/. them. 2008). Despite numerous surveys throughout Gauteng, only four specimens of the three Responses of reptiles to habitat modification species - two of C. aenea and two of H. dor- have been linked to numerous factors that salis (see Whittington-Jones et at. 2008; D. include the type ofdisturbance involved (Jones Koen, unpub!. data) - have been collected since et al. 2000); changes in microhabitat availabil- 2000. The rarity of these species coupled with ity (James & M'Closeky 2003; Goode et al. the rapid increase in human population density 2005); changes in the rates ofpredation follow- in the province suggests that the pressure on ing modification (Reinert 1993); changes in the these grassland species is increasing. thermal properties of the habitat (Lillywhite 1987) and intrinsic characteristics such as the Recognising the pressure on the natural envi- species' dispersal and recolonisation abilities ronment and the links between reptile diversity (Twigg & Fox 1991). Due to their limited and habitat structure (Mushinsky 1985; Block mobility, reptiles are susceptible to habitat et at. 1998; Cavitt 2000; Woinarski & Ash fragmentation and transformation (Webb & 2002; Read 2002; James 2003; Maritz and Shine 1997). Continued habitat transformation Alexander 2007; Masterson et at. 2008), we and fragmentation ofreptile populations within investigated the effects of land-use and habitat a landscape can lead to extirpation, but the type on the species richness and abundance of responses ofparticular species to habitat modi- reptile assemblages in three habitat types (two fication are not easily predicted and often natural and one modified) in the grasslands of require explicit testing or monitoring. Gauteng, South Africa. The primary aim ofour investigation was to assess the impacts of his- Gauteng lies at the interface between the torical cultivation on the local reptile assem- Savanna and Grassland biomes of South blages in these threatened grasslands. Africa. Grasslands in South Africa are under extreme pressure and formally protected areas Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 account for only 2%of the biome's total area MATERIALS AND METHODS (Bredenkamp 2002). Threats to the Grassland biome include: (1) high agricultural potential, Study Site and Reptile Survey.-Suikerbosrand (2) high mining potential, (3) habitat fragmen- Nature Reserve (260 30" S; 28 0 15" E) is locat- tation and (4) the high suitability for the inva- ed approximately 40 km south of sion ofalien plant species (Driver et al. 2005). Johannesburg, Gauteng Province, South Africa, Consequently, the Grassland biome has the and incorporates the major portion of the highest proportion of threatened ecosystems Suikerbosrand, a high lying plateau named for

72 MASTERSON ET AL.- Reptiles in historically-transformed grasslands

CJ OIdL....s

CJ S.N.R. Exteasiolo

_ Soikeftosraad N.R.

o 4 8

Figure I. A map ofthe study site, the positions ofthe nine trap arrays in relation to each other and the three historically cultivated areas (numbered polygons) sampled. The location of Suikerbosrand Nature Reserve in Gauteng, South Africa is shown (inset).

the abundant Highveld Protea (Protea cafJra). We surveyed reptiles on the extension of Originally proclaimed in 1973, Suikerbosrand Suikerbosrand Nature Reserve between 1 Nature Reserve was enlarged by the purchase December 2005 and 20 April 2006 using trap of 6 936 ha of adjacent mixed agricultural arrays. Nine trap arrays were installed in three lands in 2005. As a result of the acquisition, habitat types i.e., three clusters ofsites (Fig. 1). Suikerbosrand Nature Reserve encompasses Trap clusters were situated 1 976 ± 550 m from each other (min = 1 398 m; max = 2495 m), 18587 ha and is the largest grassland reserve in while trap arrays within clusters were Gauteng. Altitude on the reserve ranges from 518 ± 204 m apart (min = 223 m; max 1 545 to 1 917 m a.s.l. Rainfall is highly sea- = 830 m). Habitat types in each cluster were (I) sonal, with most ofthe annual mean of675 mm primary grassland with little or no rock cover Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 falling in summer, between October and March (pristine sites), (2) primary grassland with large (Shultze 1997). Vegetation on the recently- quartzite outcrops and scattered rocks (rocky acquired extension is classified as Tsakane sites) and (3) secondary grasslands that were Clay Grassland (Mucina & Rutherford 2005). historically ploughed, cultivated and cropped Tsakane Clay Grassland is a highly threatened (modified sites). Trap arrays in the modified vegetation type with 63 % of the vegetation habitats were placed between 50 to 100 m from type irreversibly transformed (Mucina & the habitat edge. The layout of the trap arrays Rutherford 2006). used in this field survey has been described

73 AFRICAN JOURNAL OF HERPETOLOGY 58(2) 2009

elsewhere (Maritz et al. 2007). In short, each ages) were arcsine transformed prior to data array consisted of 36 m of plastic drift fence, analysis. five pitfall traps and eight mesh funnel traps installed in a closed-cross configuration. Traps Data Analysis.-We compared tht; mean were checked daily and all captured reptiles species richness and mean number of reptiles were identified to species level and released at captured in each habitat type and each cluster the site ofcapture. Lizard species were marked using one-way Analysis ofYariance (ANaYA). using a trap-specific toe-clip code to allow us We used General Linear Modelling (GLM) to to identifY dispersal events between trap arrays. test the significance and compare the explana- species were not marked. Nomenclature tory power of (1) habitat type, (2) cluster, (3) is from Branch (1998), although we use the mean vegetation height, (4) mean vegetation generic assignment of Fitzinger cover and (5) the total number ofreptiles cap- 1843 instead of Mabuya Fitzinger 1826 (see tured on the observed species richness (Sobs) at Bauer 2003). each trap array. Factors with a significant effect on the total number of species recorded at each The disturbance histories of the three cultivat- site were then combined using forward selec- ed areas differ from each other and the sur- tion to test the significance oftheir interactions. rounding grassland. Cultivated area 1 was We also compared the relative species richness ploughed in 2005 just prior to our survey in i.e., species richness per number ofindividuals order to allow for the re-seeding of indigenous captured, and species density i.e., species rich- species at the site. Cultivated area 2 has not ness per sample, of each habitat type and clus- been ploughed since at least 2002 and possibly ter using sample-based rarefaction (Gotelli & even 2000, while sunflowers were last planted Colwell 2001). Sample-based rarefaction on cultivated area 3 during the 2002/2003 curves were produced using EstimateS version growing season. Unfortunately, accurate fire 7.5 (Colwell 2005) with a sample defined as the records for the extension exist only from 2004 number of reptiles captured per group of traps onwards when the area was included in i.e., habitat type or cluster, per day (N = 141 Suikerbosrand Nature Reserve, but it is likely samples). that fire frequency in the cultivated areas was reduced by the pre-spring ploughing and seed- There is some disagreement as to the specific ing. The majority of natural fires at method by which confidence intervals should be Suikerbosrand Nature Reserve are started by used to confirm differences between rarefaction lightning during the early spring rains in curves. Colwell et al. (2004) tested for the over- September and October. lap of confidence intervals in their comparison of old growth and secondary growth forests. In Vegetation Measures.-We measured two contrast, Magurran (2004) analysed only the structural characteristics ofthe vegetation with- position of the mean value of the smaller rar- in 400 m2 of each trap array using a grid of 20 efaction curve relative to the confidence interval Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 2 quadrats (size = 1m; as per Masterson et al., of the larger curve in her comparison of 2008). Yegetation height was measured at each Drosophila species richness between two sur- ofthe four corners of each quadrat; vegetation veys. The choice of method is strongly influ- cover was estimated as the proportion of soil enced by whether or not the two assemblages obscured from view by the vegetation within being compared can be assumed to be equivalent the quadrat. Mean vegetation height and mean in their species composition. If the two assem- vegetation cover were calculated for each trap blages can be assumed or are known to be equiv- array. Measures of vegetation cover (percent- alent, then the exclusion of the mean of the

74 MASTERSON ET AL.- Reptiles in historically-transfonned grasslands

Table I. A summary of the total number of individual reptiles captured per habitat type and per species between December 2005 and April 2006. Recaptures ofthe toe-clipped lizard species are shown in brackets.

Family Species Number Habitat type captured ModIfIed Pnstme Rocky

Lizards Agamidae Agama aculeata distanti 9 (2) 2 7 (2) Cordylidae Cordylus vittifer 4 (0) 3 Gerrhosauridae Gerrhosaurus flavigularis 116 (40) 7 64 (30) 45 (10) Lacertidae Nucras lalandii 4 (I) 4 (I) Scincidae Panaspis walbergii 33 2 6 25 Trachylepis capensis 130 (20) 59 (8) 50 (9) 21 (3) Trachylepis varia 26 (I) 3 5 (I) 18 Gekkonidae Pachydactylus capensis 2 2 Typhlopidae Typhlops bibronii Atractaspididae Aparallactus capensis 5 32 Atractaspis bibronii I I Colubridae Crotaphopeltis hotamboeia I I Dasypeltis scabra 17 4 85 Lamprophis aurora 4 I I2 Lamprophis capensis 2 2 Lycophidion capense 20 I 8 II Psammophis crucifer 17 2 96 Psammophylax rhombeatus 13 2 10 I Pseudaspis cana 3 2 I Elapidae Hemachatus haemachatus 8 3 23 Viperidae Causus rhombeatus 15 3 4 8 Bitis arietans arietans 12 8 4

Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 All species 443 (64) 89 (8) 196 (43) 158 (13) smaller curve from the confidence interval ofthe With no historical data on community compo- larger curve is sufficient evidence of a signifi- sition available to us, we assumed that the rep- cant difference in the species richness ofthe two tile assemblages varied between the three habi- assemblages. Iftwo assemblages are assumed to tat types and used the more conservative test in be different in their species composition then the habitat comparisons. Consequently, we used more conservative test, using the overlap of the the non-overlap of the confidence intervals at confidence intervals, is most appropriate. the largest sample size of the shorter rarefac-

75 AFRICAN JOURNAL OF HERPETOLOGY 58(2) 2009

tion curve to identify significant differences verse that were observed during sampling between two curves. We also formally tested (Soberon et al. 2007). Estimates of species our assumption regarding the variation in rep- richness for the three habitat types were gener- tile assemblages by comparing the similarity of ated using the nonparametric Chao1 and Cha02 the reptile assemblages recorded in each of the species richness estimators (Chao 1984; three habitat types and clusters using the Colwell 2005). The equations for the two esti- Analysis of Similarity (ANOSIM) sub-routine mators are functionally equivalent, but Chao1 in Primer 5 (Clarke & Gorley 2001). Due to the uses abundance frequencies i.e. the number of right skew in the capture counts ofthe different individuals of each species captured, while species at each site, we log-transformed our Cha02 uses incidence frequencies i.e., the num- data before calculating the Bray-Curtis ber of samples in which each species is Similarity between each pair ofsites. Although observed. The terms 'singleton' and 'doubleton' skewness was not completely removed by this refer to species that are represented by only one transformation it was greatly reduced. or two individuals in the total sample respec- tively, while the terms 'unique' and 'duplicate' Lastly, we evaluated the 'completeness' of our refer to species that are observed in only one or reptile sampling. In a sampling context, com- two samples respectively. The estimators use pleteness is evaluated as the percentage of the ratio ofsingletons to doubletons or uniques species estimated to occur in the sampling uni- to duplicates to estimate the number of species A 18 B 18 16 16 14 14 '" '" ..'"c 12 ..'"c 12 ..c ..c u ii1 10 ii2 10 8 8 'u..'" .!!'" 8- 6 6 ell ell 4 4 2 2 0 0 2 3 Pristine Rocky Modified Cluster Habitat type C 100 D 100 90 90 80 80 70 70 60 60 0... 50 I 50 u 40 u.. 40

Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 30 30 20 20 10 10 0 0 2 3 Pristine Rocky Modified Cluster Habitat Type

Figure 2. Mean reptile species richness for each cluster (A), and in each habitat (B), and mean number of reptile captures for each cluster (C) and each habitat (D), from trap array surveys in a South African grass- land. Error bars indicate 95 % confidence limits.

76 MASTERSON ET AL.- Reptiles in historically-transfonned grasslands

Table 2. Observed (Sobs) and estimated species richness, with 95 % confidence intervals, for each ofthe three habitat types at two sample sizes (n). N refers to the total number ofindividuals captured in each habitat type during the sampling and is shown in Table I for each habitat type.

Habitat Type Sobs ±1SD SChaol 95%CI SChao2 95%CI Mean Lower Upper Mean Lower Upper

n = N individuals Pristine 19± 0.70 19.20 19.01 23.06 19.20 19.01 23.04 Rocky 18 ± 2.15 21.33 18.54 38.47 21.31 18.54 38.34 Modified 12 ± 0.62 12.20 12.01 16.06 12.17 12.01 15.56

n = 89 individuals Pristine 16.16 ± 0.97 18.94 16.70 32.26 18.92 16.70 31.90 Rocky 15.24 ± 1.98 19.63 16.16 38.35 19.25 16.07 36.66 Modified 12.00 ± 0.62 12.20 12.01 16.06 12.17 12.01 15.56 that were not detected during the sampling. crucifer, Gerrhosaurus flavigularis, Panaspis walbergii, Trachylepis capensis and We calculated two estimates of total species Trachylepis varia. Trachylepis capensis was richness for the three habitat types using the most frequently captured species at sites in EstimateS (Colwell 2005). The first estimate the modified habitat, while G flavigularis was was calculated using the maximum sample the most frequently captured species at pristine sizes available for each habitat type i.e., indi- sites. viduals captured and samples taken, while the second estimate was calculated when the num- Sixty-four lizards were recaptured during the ber of individuals captured in the pristine and survey (Table I). Forty-three of the recaptures rocky habitat types equalled that of the total (67.2 %) were recorded at pristine sites, 13 number captured in the modified habitat. (20.3 %) were recorded at rocky sites and eight (12.5 %) were recorded at modified sites (Table I). Gerrhosaurusflavigularis was the most fre- RESULTS quently recaptured species during the survey period, but T. capensis was the only species During the four months of trapping, we cap- that was recaptured in all three habitat types tured 443 reptiles comprising 8 lizard and 14 and the only species to be recaptured at modi- snake species (see Table 1 for details). Ten of fied sites (Table I). There was no indication of the 22 species we captured were not captured in dispersal between trap arrays during the survey the modified habitat. Three species of snake, as all lizards were recaptured at the site oftheir Crotaphopeltis hotamboeia, Atractaspis initial capture. bibronii and 1jlphlops bibronii, were captured only once during the survey and only in the There was no significant difference in the total Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 rocky habitat. Two species of lizard, Nucras number ofspecies recorded at each ofthe three

lalandii and Pachydactylus capensis, were clusters (ANOYA: F2•6 = 0.47; P = 0.64; Fig. recorded on multiple occasions from only a sin- 2A) but significantly fewer species were cap- gle habitat type, i.e. pristine sites. Eight species tured at modified sites than at the pristine or

(four snake and four lizard species) were cap- rocky sites (ANOYA: F2•6 = 11.63; P = 0.009; tured twice or more in all three habitat types: Tukey HSD Post-hoc: Pristine vs. Rocky, P = Causus rhombeatus, Dasypeltis scabra, 0.74; Pristine vs. Modified, P = 0.01; Rocky vs. Hemachatus haemachatus and Psammophis Modified, P = 0.02; Fig. 2B). We found no sig-

77 AFRICAN JOURNAL OF HERPETOLOGY 58(2) 2009

A

25 - Pristine - Modified

20

.l!i·• 15

.!! · 10 .i"

5

0

0 50 100 150 200

Indlvld...... Captured B

25 - Rocky - Modified

20

: ·<: 15 u it! .!! ·¥ 10 a. UJ

5

0

0 50 100 150 200

Indlvld...... Captured

Figure 3. Species richness per number ofindividuals for the pristine and modified habitat types (A) and rocky

Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 and modified habitat types (B) sampled between December 2005 and April 2006. Error bars indicate the 95 % confidence intervals for each curve, calculated by EstimateS version 7.5 (Colwell 2005). Statistically sig- nificant differences between the curves were determined by the non-overlap of confidence intervals at the point ofcomparison. nificant difference in the total number of rep- Results of our sample-based rarefaction analy- tiles captured at sites in each cluster (ANaYA: sis were similar but not identical to the results

F2•6 = 0.31; P = 0.74; Fig. 2C) or each habitat of our ANaYA. As with the ANaYA, differ-

type (ANaYA: F2•6 = 3.79; P = 0.09; Fig. 2D). ences in reptile species richness of the three

78 MASTERSON ET AL.- Reptiles in historically-transfonned grasslands

A

25 - Pristine - Modified

20

i .l! 15 iii!" 11· ·Q. 10 '"

5

0

0 50 100 150 sample. B

25 - Rocky - Modified

20

: l: 15 .I: 6!· :

l!Q. 10 '"

5

0

0 50 100 150 sample.

Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 Figure 4. Species richness per number of samples for the pristine and modified habitat types (A) and rocky and modified habitat types (8) sampled between December 2005 and April 2006. Error bars indicate the 95 % confidence intervals for each curve, calculated by EstimateS version 7.5 (Colwell 2005). Statistically sig- nificant differences between the curves were determined by the non-overlap of confidence intervals at the point ofcomparison. clusters were not significant, while pristine modified sites, but unlike the result of the sites had significantly greater species richness ANOYA, the difference was not statistically than modified sites (Fig. 3A). More reptile significant due to confidence interval overlap species were captured at rocky sites than at (Fig. 3B). Modified sites had both the lowest

79 AFRICAN JOURNh OF HERPETOLOGY 58(2) 2009

observed species richness and the narrowest 95 least different in species composition and abun- % confidence interval ofthe three habitat types dance (ANOSIM: R = -0.11, Npennutations = 10, P (Sobs ± 1.96 SD = 12 ± 1.11 species). Species = 0.60). density, the number of species recorded per sample, at the modified sites was significantly Estimates of species richness for each of the lower than the species density of both the pris- three habitat types are shown in Table 2. The 95 tine (Fig. 4A) and the rocky sites (Fig. 4B). % confidence interval for the total estimated Rocky sites showed the greatest heterogeneity species richness in the modified habitat ranged in sample richness, as seen by the width of the from 12 to 16 species (Table 2), suggesting that confidence interval for rocky sites (Sobs ± 1.96 the sampling of the species found in the modi- SD = 18 ± 6.09 species). Species density varied fied habitat was between 75 and 100 % com- insignificantly between the three clusters of plete (mean = 98.36 %). At the equivalent level trap arrays. of sampling effort i.e., 89 individuals, the con- fidence intervals of the estimated species rich- Species richness was well explained by habitat ness in the pristine and rocky habitats ranged type (GLM: F(2.6) = 11.62, P = 0.009, R' = 0.79) from 17 to 32 species and 16 to 38 species and the total number of reptiles captured respectively, suggesting the presence of nearly (GLM: F(I.7) = 18.94, P = 0.003, R' = 0.73) yet twice as many species as had been observed up to that point in both habitats (Table 2). With the the bivariate model including habitat type and inclusion of additional individuals, the upper the total number of captures indicated a non- bound of the confidence intervals for the pris- significant effect of the total number of cap- tine habitat decreased to 23 species, but tures on the observed species richness. Mean remained at 38 species for the rocky habitat vegetation cover, mean vegetation height and (Table 2). Consequently, the estimated sample cluster had no significant effect on the completeness for the pristine habitat ranges observed species richness recorded at each site from 83 and 100 % (mean = 98.96 %), while (GLM: P> 0.05 in all univariate models). the sample completeness of the rocky habitat ranges from 47 and 100 % (mean = 84.39 %). The species composition ofreptile assemblages varied by habitat type (ANOSIM: Global R = 0.325, Npennutations = 280, P = 0.075), but not by DISCUSSION cluster (ANOSIM: Global R = -0.029, Npennutations = 280, P = 0.53). Pairwise compar- The results of our two analyses indicate that isons of species similarity indicated a strong modified, secondary grassland in historically difference in the species composition of the cultivated areas supports fewer species than the modified and the rocky habitats (ANOSIM: R equivalent primary grassland. Differences in = 0.67, Npennutations = 10, P = 0.1 0) but little dif- species richness were not significantly influ- ference between the modified and pristine habi- enced by the spatial location of each site clus- Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 tats (ANOSIM: R = 0.33, Npennutations = 10, P = ter or the structure of the vegetation in and 0.30). Although P> 0.05 in the modified-rocky around each trap array. As with species rich- pairwise comparison, the grouping of sites by ness, we found no evidence for an effect of habitat type produced as close a probability to cluster on species composition. Our results 0.05 as could be achieved given the number of indicate that habitat transformation via land- available permutations i.e., only one of the 10 use change has had a detectable, negative permutations of R* was greater than R. Trap impact on the species richness ofthe local rep- sites in the pristine and rocky habitats were the tile assemblages in the extension of

80 MASTERSON ET AL.- Reptiles in historically-transfonned grasslands

Suikerbosrand Nature Reserve. Estimates of basking, refuge and foraging e.g., Cordylus vit- species richness indicated that both the pristine tifer and Aparallactus capensis (Table I). We and modified habitats were thoroughly sampled found no difference in the structure of the veg- during our survey, while rocky habitat was sat- etation between the traps placed in the three isfactorily, yet not as thoroughly, sampled. We habitat types, which suggests that reptile diver- also note that our results may even be consid- sity does not solely depend on physical charac- ered a conservative assessment of the impacts teristics of the vegetation, but may instead be of habitat transformation, given that our trap linked to properties of the primary grassland arrays were close to the edges of the modified communities that we did not assess e.g., per- habitat and may have been affected by edge centage natural vegetation cover. effects. Measures ofassemblage changes in response to Reptile assemblages are known to vary as the habitat transformation are influenced by the land-use ofan area varies (Castellano & Valone strength and or proportion of positive, neutral 2006; Santelmann et al. 2006). Land-use and negative responses of the species within change leads to changes in vegetation cover, the assemblage. Certain species may benefit which can influence other ecological factors from the sharp edges created by disturbance such as predation rates (Castellano & Valone e.g., lizards that use the edge to shuttle between 2006). Santelmann et al. (2006) used weighted sunlight and shade (Duelli 1997; Fabricius et habitat associations to model the impacts of af. 2003), while other species may suffer as a future agricultural land-use scenarios on the result of the reduced complexity of the habitat wildlife in Iowa, USA. They found that suitable or vegetation (Maissoneuve & Rioux 200 I; habitat for reptile species was most negatively Jobin et af. 2004) or a reduction in natural veg- affected by a scenario in which the landscape etation (Lindenmayer et al. 2005; Hodgkison et was zoned for profitable agricultural produc- al. 2007). The degree to which one is able to tion. Santelmann et af. (2006) also found that characterise the response of an assemblage is the response of reptiles to the future land-use determined by the completeness of the sam- scenarios differed from that of other verte- pling of that assemblage. Species richness esti- brates, which suggests that reptiles may need to mates for the three habitat types in our surveys be given special consideration in land-use plan- supported the observed differences in species ning scenarios and that the use of faunal surro- richness between them. Estimates of species gates may not be appropriate for reptiles in all richness that utilise the ratio of ecosystems. singletons/unique to doubletons/duplicates are best considered as a lower bound of the total The reptile assemblage in the northern exten- species richness (Mao & Colwell 2005). There sion of Suikerbosrand Nature Reserve appears have also been concerns raised regarding the to be sensitive to the changes that have resulted truthfulness of the 95 % confidence intervals, from previous agricultural land-use. The most as it can be shown that the number ofundetect- Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 palpable consequence of agricultural land-use ed species may be very large without influenc- is the homogenisation of the habitat, both in ing the width of the confidence interval (Mao terms of vegetation monocultures and micro- & Colwell 2005). For this reason we did not habitat loss such as resulted from the removal compare the estimated species richness for of rocks prior to ploughing (D. Koen pers. each of the three habitat types, we simply used comm.). The effect of rock removal appears to the estimate to provide us with a measure of have led to a reduction in the number of cap- sample completeness. Mean estimates of sam- tures of species that utilise rock outcrops for ple completeness for the three habitats were

81 AFRICAN JOURNAL OF HERPETOLOGY 58(2) 2009

exceptionally high for the pristine and modified modified sites, the majority of the T. capensis habitats, and more than satisfactory for the individuals captured during the survey were rocky habitat. The combination of high sample captured on the modified sites. For the majori- completeness and the observed differences in ty of species, the number of captures recorded species richness of the three habitat types pro- at the pristine or rocky sites was greater than or vide strong evidence of a significant decline in equal to the number captured on the modified the species richness of the reptile assemblage sites. The number of captures of T. capensis at following habitat transformation associated the modified sites may have been caused by an with cultivation. increase in the frequency or distance of move- ments by individuals and an associated increase Nucras lalandii (Lacertidae) appears to be in the probability of being captured. highly sensitive to land-use change. All of the Alternately, T. capensis may actually be more individuals of N. lalandii captured in our sur- abundant at modified sites than in the pristine vey were captured at a single locality in a low, or rocky habitats. In the pristine and rocky closed grassland (sensu Edwards 1983), with habitats, T. capensis may be affected by inter- no captures recorded at the nearby modified specific competition with Gerrhosaurus jlav- site. Despite extensive surveys throughout igularis (Gerrhosauridae), which fills a similar Gauteng between 2000 and 2008, all five ofthe ecological niche as T. capensis, and which post-2000 localities for N. lalandii records are occurs in greater numbers in these two habitat situated in primary grassland on the extension types than in the modified habitat. Whatever of Suikerbosrand Nature Reserve the reasons, the response of T. capensis to habi- (Whittington-Jones et af. 2008). The popula- tat transformation was exceptional among all tion of N. lalandii in Suikerbosrand Nature the reptile species captured during our survey. Reserve is isolated from populations to the north and east by approximately 100 - 200 km In systems where the natural vegetation plays (Jacobsen 1989; Jacobsen, 1995), and the an important role in faunal diversity, manage- species' absence in historically cultivated areas ment actions that improve the condition of the suggests that the species may be declining in vegetation can have positive spin-ofTs for the Gauteng's highly-transformed landscape. associated fauna (Castellano & Valone 2006), Surprisingly, N. lalandii has not been recorded and vice versa (Santelmann et al. 2006). within the 1973 boundary of Suikerbosrand Currently the modified sites are dominated by Nature Reserve despite surveys in 2004, 2005, weedy species, e.g. Bidens pi/osa (blackjack), 2006 and 2007 (Koen & du Toit 2007; but a reseeding programme currently being Masterson 2007; Masterson et af. 2008) and the implemented in the historically-cultivated reasons for this absence are currently unknown. areas ofthe extension to Suikerbosrand Nature The persistence of N. lalandii in Gauteng may Reserve is showing promising results in restor- depend on the appropriate protection and man- ing the natural vegetation in these transformed agement of primary grassland patches where areas (D. Koen, unpub!. data). Simultaneous Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 the species occurs. monitoring ofthe vegetation and reptile assem- blages on the extension of Suikerbosrand In contrast with the majority of reptile species Nature Reserve would be useful in order to observed during our study, Trachylepis capen- determine the effects ofvegetation recovery on sis (Scincidae) appears to benefit from habitat reptile species richness and composition. transformation associated with cultivation. Even with the reduced species richness and In conclusion, we recommend that land-use reduced number of reptiles captured at the planning in Gauteng emphasise the need for

82 MASTERSON ET AL.- Reptiles in historically-transfonned grasslands

areas of connected, untransfonned habitat in BRANCH, WR. 1998. Field Guide to Snakes and Other order to mitigate the negative effects of habitat Reptiles of Southern Africa (3rd ed.). Struik, Cape Town. transfonnation on reptile assemblages. In the BREDENKAMP, G.J. 2002. The Grassland Ecoregion. In: Le highly transfonned landscape of Gauteng, the Roux, J. (Ed.) The Biodiversity of South Africa 2002: protection of untransfonned habitat will play Indicators, Trends and Human Impacts. Struik, Cape an important role in the persistence of species Town. CASTELLANO, M.l. & T.J. VALONE. 2006. Effects of live- that are sensitive to disturbance e.g., N. stock removal and perennial grass recovery on the lalandii. Where patches of primary grassland lizards of desertified arid grassland. J. Arid Environs. do not occur, disturbance management and 66: 87-95. habitat/vegetation restoration may be required CAVITT, J.F. 2000. Fire and a tallgrass prairie reptile com- munity: Effects of fire on relative abundance and sea- to protect and/or restore key elements of the sonal activity. J. Herpetol. 34: 12-20. habitat. Currently it is not known whether the CHAO, A. 1984. Nonparametric estimation ofthe number of regeneration ofnatural vegetation will be sufJi- classes in a population. Scand. J. Stat. II: 265-270. cient to improve reptile diversity in the affect- CLARKE, K.R. & R.N. GORLEY. 2001. PRIMER v5: User Manual and Tutorial. Plymouth, PRJMER-E Ltd, UK, ed grasslands but regular monitoring ofthe rep- p.91. tiles at sites with regenerating vegetation is rec- COLWELL, R.K. 2005. EstimateS: Statistical estimation of ommended. species richness and shared species from samples. Version 7.5. User's Guide and application published at: http://purl.oclc.org/estimates. COLWELL, R.K., C.X. MAO & J. CHANG. 2004. Interpolating, ACKNOWLEDGEMENTS extrapolating and comparing incidence-based species accumulation curves. Ecology. 85: 2717-2727. We thank Daniel Koen and Neville Green for DEPARTMENT OF ENVIRONMENTAL AFFAIRS AND TOURISM. 2005. South Africa Country Study. National assisting us with the site selection and installation Biodiversity Strategy and Action Plan pp 9 and 74. ofthe trap arrays and thank Liesl du Toit for pro- DRIVER, A., K. MAZE, M. ROUGET, A.T. loMBARD, J. NEL, viding infonnation on the site histories ofthe culti- l.K. TURPIE, R.M. COWLING, P. DESMET, P. GOODMAN, J. vated areas. We also thank the Gauteng HARRIS, Z. JONAS, B. REYERS, K. SINK & T. STRAUSS. 2005. National Spatial Biodiversity Assessment 2004: Department ofAgriculture and Rural Development priorities for biodiversity conservation in South Africa. for pennission (Pennit # Zl 1314) to conduct this Strelitzia 17. South African National Biodiversity survey and two anonymous reviewers whose com- Institute, Pretoria. ments and suggestions helped us to improve the DUELLI, P. 1997. Biodiversity evaluation in agricultural landscapes: an approach at two different scales. Agric., quality ofthis manuscript. All methods and proto- Ecos. & Environ. 62: 81-91. cols were screened and approved by the EDWARDS, D. 1983. A broad-scale structural classification of Ethics Committee of the University of the vegetation for practical purposes. Bothalia 14: 705-712. Witwatersrand (pennit # 2005/67/1). FABRICIUS, C., M. BURGER & P.A.R. HOCKEY. 2003. Comparing biodiversity between protected areas and adjacent rangeland in xeric thicket, South Africa: arthropods and reptiles. J. Arid Environs. 40: 392-403. LITERATURE CITED GOODE, M.J., WC. HORRACE, M.J. SREDL & J.M. HOWLAND. 2005. Habitat destruction by collectors

Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 BAUER, A.M. 2003. On the identity of Lacerta punctata associated with decreased abundance of rock-dwelling Linnaeus 1758, the type species ofthe genus Euprepis lizards. BioI. Cons. 125: 47-54. Wagler 1830, and the generic assignment of Afro- GOTELLI, N.J. & R.K. COLWELL. 2001. QuantifYing biodiver- Malagasy . Afr. J. Herpetol. 52: 1-7. sity: procedures and pitfalls in the measurement and com- BLOCK, W.M., M.L. MORRISON & P.E. SCOTT. 1998. parison ofspecies richness. Ecol. Lett. 4: 379-391. Development and evaluation ofhabitat models for her- HODGKISON, S.c., J. HERO & l. WARNKEN. 2007. The conser- petofauna and small mammals. For. Sci. 44: 430-437. vation of suburban golf courses in rapidly urbanising BRANCH, WR. 1988. South African Red Data Book - region of Australia. Landscape & Urban Planning. 79: Reptiles and Amphibians. Pp. 242. South African 323-337. National Scientific Projects Report 151. JACOBSEN, N.H.G. 1995. The Herpetofauna of Gauteng

83 AFRICAN JOURNAL OF HERPETOLOGY 58(2) 2009

Province - localities and distribution maps. Report. Chief fauna in a South African grassland. Applied Herpetol. 5: Directorate ofNature and Environmental Conservation. 129-143. JACOBSEN, N.H.G 1989. A herpetological survey of the MUCINA, L. & M.e. RUTHERFORD. 2005. Vegetation Map of Transvaal. Unpubl. Ph.D. thesis. University of Natal. South Africa, Lesotho and Swaziland. National Botanical South Africa. Institute, South Africa. JAMES, e.D. 2003. Responses ofvertebrates to fenceline con- MUCINA, L. & M.e. RUTHERFORD (Eds.). 2006. Vegetation trasts in grazing intensity in semi-arid woodlands ofeast- map of South Africa, Lesotho and Swaziland: an illus- ern Australia. Austral Ecol. 28: 137-151. trated guide. Strelitzia 19, South African National JAMES, S.E. & R.T. M'CWSEKY. 2003. Lizard microhabitat Biodiversity Institute, Pretoria. and fire fuel management. BioI. Conserv. 114: 293-297. MUSHINSKY, H.R. 1985. Fire and the Florida sandhill herpeto- JOBIN, B., L. BELANGER, e. BOUTIN & e. MAISONNEUVE. 2004. faunal community: with special attention to responses of Conservation of agricultural riparian strips in the Boyer Cnemidophorus sex/ineatus. Herpetologica 41: 333-342. river watershed, Quebec (Canada). Agric., Ecosys. & READ, J. 2002. Experimental trial ofAustralian arid zone rep- Environ. 103: 413-423. tiles as early warning indicators ofovergrazing by cattle. JONES, B., S.F. Fox, D.M. LESLIE, D.M. ENGLE & R.L. Aus. Ecol. 27: 55-66. LocHMILLER. 2000. Herpetofaunal responses to brush REINERT, H.K. 1993. Habitat Selection in Snakes. In: Seigel R. management with herbicide and fire. J. Range. Man. 53: A., Collins J. T. & Novak S.S. (Eds.), Snakes: Ecology 154-158. and Evolutionary Biology. McGraw-Hill Company, New KOEN, D. & L. DU TOIT. 2007. Survey ofthe herpetofauna of York. the Suikerbosrand Nature Reserve and extension for SANTELMANN, M., K. FREEMARK, J. SIFNEOS & D. WHITE. 2006-2007. Report. Chief Directorate of Nature 2006. Assessing effects of alternative agricultural prac- Conservation. tices on wildlife in Iowa, USA. Agric., Ecosys. & LILLY WHITE, H.B. 1987. Temperature, energetics, and physio- Environ. 113: 243-253. logical ecology. In: Seigel R. A., Collins J.T., Novak S. S. SHULZE, R.E. 1997. South African Atlas ofAgrohydrology and (Eds.), Snakes: Ecology and Evolutionary Biology. Climatology. Water Research Commission, Pretoria, Macmillan Publishing Company, New York, pp. 422-477. South Africa. Report IT 82/96. LINDENMAYER, D.B., J. FISHER & R.B. CUNNINGHAM. 2005. SMART, R., M.J. WHITING & W. TWINE. 2005. Lizards and Native vegetation cover thresholds associated with landscapes: integrating field surveys and interviews to species responses. BioI. Cons. 124: 311-316. assess the impact ofhuman disturbance on lizard assem- MAGURRAN, A. 2004. Measuring biological diversity. Blackwell Publishing, Oxford. blages and selected reptiles in a savanna in South Africa. BioI. Cons. 122: 23-31. MAISONNEUVE, e. & S. RIoux. 200 I. Importance of riparian habitats for small mammal and herpetofaunal communi- SOBERON, J., R. JIMENEZ, J. GOLUBOV & P. KOLEFF. 2007. ties in agricultural landscapes ofsouthern Quebec. Agric., Assessing completeness of biodiversity databases at dif- Ecosys. & Environ. 83: 65-175. ferent scales. Ecography 30: 152-160. MAO, e.X. & R.K. COLWELL. 2005. Estimation of species STATISTICS SOUTH AFRICA (2008) Mid-year population esti- richness: mixture models, the role of rare species, and mates 2008. Statistical release P0302. inferential challenges. Ecol. 86: 1143-1153. TWIGG, L.E. & B.J. Fox. 1991. Recolonization ofregenerating MARITZ, B. & G1. ALEXANDER. 2007. Herpetofaunal utilisa- open forest by terrestrial lizards following sand mining. tion ofriparian buffer zones in an agricultural landscape Aus.1. Ecol. 16: 137-148. near Mtunzini, South Africa. Afr. 1. Herpetol. 56: 163- WEBB, J.K. & R. SHINE. 1997. A field study ofspatial ecology 170. and movements of a threatened species, Hoplocephalus MARITZ, B., GP.R. MASTERSON, D. MACKAY & GJ. bungaroides. BioI. Cons. 82: 203-217. ALEXANDER. 2007. The effect offunnel trap type and size WHITTINGTON-JONES, C., S. WEST, A. MATABANE, R. KOKO, W. ofpitfall trap on trap success: implications for ecological MOLABA, 1. MOTSAMAI, 1. MAKOLA, A. MPHUTI & S. field studies. Amphib.-Rept. 28: 321-328. NDZHUKULA. 2008. The herpetofauna of Gauteng: distri- MASTERSON, GP.R. 2007. Testing the effects ofhabitat modi- bution and status of reptiles. Report Volume I. Chief Downloaded by [The Library, University of Witwatersrand] at 06:19 01 October 2012 fication on the reptiles of a South African grassland. Directorate ofNature Conservation Unpubl. MSc thesis. University of the Witwatersrand, WOINARSKI, J.e.Z. & A.J. ASH. 2002. Responses of verte- Johannesburg, South Africa. brates to pastoralism, military land use and landscape MASTERSON, GP.R., B. MARITZ & GJ. ALEXANDER. 2008. positions in an Australian tropical savanna. Aus. Ecol. 27: Effect offire history and vegetation structure on herpeto- 311-323.

Received: 14 April 2009; Final acceptance: 29 July 2009

84