We sampled the seasonal diversity of rodents and shrews in two neighboring game reserves in the savannah biome and assessed the influence of competition and abiotic processes on species composition patterns at habitat and reserve scales. We used null models and multivariate analyses to test whether the observed patterns differed significantly from those expected by chance. Species richness estimators indicated that rodent and shrew inventories were relatively complete. Rarefaction curves showed that small mammal species richness was higher at the Mkhuze Game Reserve than the Kube Yini Game Reserve, though it was highest in the Acacia woodlands. Rodent diversity was highest during winter. Rodent and shrew species composition was significantly correlated with grass height, tree density, and ground cover. Furthermore, rodent and shrew assemblages were nested at the reserve scale. Only nestedness at the broadest scale (both reserves, both seasons combined) was significantly correlated with landscape and microhabitat variables. Conversely, there was little evidence that competition influenced rodent and shrew species composition. Results showed that species composition patterns of small mammals in the savannah biome are related to abiotic processes operating at local and regional scales, suggesting that microhabitat characteristics and regional processes like immigration and extinction drive assemblage structure.
Patterns that characterize assemblages of coexisting species in space and time can provide clues to underlying mechanistic processes. Biotic aspects (e.g., competition, predation, and coevolution) and abiotic processes (e.g., resource availability) may regulate species community assembly and lead to distinctive, non-random species composition patterns in local assemblages (Weiher and Keddy 2001). For example, Diamond (1975) predicted that if competition is important during assembly, certain species should co-occur less than expected by chance, creating checkerboard distributions. Similarly, only certain combinations of all possible combinations of species present in the regional pool should be observed in local assemblages. If abiotic processes are found to be predominant during assembly, different non-random species composition patterns can be predicted. For example, the nestedness hypothesis proposes that species found at species-poor sites represent subsets of species that are present at species-rich sites (Patterson and Atmar 1986). Nestedness can be produced by biogeographic processes that operate at a regional scale, such as immigration and extinction, or by habitat filtering that operates at a local scale. With over three decades of continued improvements in statistical scrutiny, non-random co-occurrence patterns consistent with predictions from biotic and abiotic hypotheses have been described in a wide range of taxa (including microorganisms, invertebrates, and vertebrates), and habitats (e.g., Patterson and Atmar 1986, Fischer and Lindenmayer 2005, Horner-Devine et al. 2007).
Rodents and shrews play important ecological roles by sustaining multiple terrestrial and aerial predators (Andersson and Erlinge 1977) and by contributing to the cycling of nutrients (Clark et al. 2005). Furthermore, they act as useful indicators of ecological integrity and can be used to predict environmental change (Avenant and Cavallini 2007). Significant nested patterns have been detected in rodent (Patterson and Brown 1991, Kelt et al. 1999, Abu Baker and Patterson 2011) and shrew (Patterson 1990) assemblages. By contrast, non-random co-occurrence patterns consistent with competition theory have been found in rodent assemblages in American deserts and in Egypt (Kelt et al. 1999, Brown et al. 2000, Abu Baker and Patterson 2011) as well as in shrew assemblages in Australian and North American temperate forests (Fox and Kirkland 1992, McCay et al. 2004).
However, in many of these studies, co-occurrence patterns are analyzed over large geographic scales that comprise heterogeneous environmental conditions (e.g., vegetation types, topography, geology, microclimate, disturbance history). Integrating heterogeneous sites in co-occurrence analyses might lead to false conclusions about species assembly, because the effects of competition and habitat filtering cannot be differentiated (Gotelli and Graves 1996). Further, most studies have focused on the influence of either biotic or abiotic processes. Such processes operate over multiple spatio-temporal scales and may overlap with each other (Lawton 2000, Schoeman and Jacobs 2008). To disentangle the processes behind species composition patterns, it is necessary to test multiple theories with the same data set that covers different habitats at varying spatial and temporal scales.
In this study, we sampled rodent and shrew diversity during wet and dry seasons in different savannah habitat types of two abutting nature reserves in South Africa. We used null models to test whether species composition patterns at habitat and reserve scales were consistent with those predicted by Diamond’s (1975) assembly rules, the niche limitation hypothesis (Wilson et al. 1987), the nestedness hypothesis (Patterson and Atmar 1986), and the habitat filtering hypothesis (Weiher and Keddy 1999).
Materials and methods
This study was conducted in Mkhuze Game Reserve (MGR; 27°35′S–27°44′S, 32°08′E–32°25′E) and the adjoining Kube Yini Private Game Reserve (KYGR; 27°42′S–27°45′S, 32°15′E–32°16′E), located at KwaZulu-Natal Province, South Africa. The two reserves have different histories. MGR was established in 1912, while KYGR was established in 1989. They also manage different species of large herbivores and predators: MGR hosts the “big five” (Du Toit et al. 2001), while KYGR only has two of them (leopards and white rhinos). The reserves fall in the savannah biome of southern Africa (Rutherford et al. 2006). The vegetation of the savannah biome is characterized by a continuous cover of perennial grasses with scattered shrubs and isolated trees (Rutherford et al. 2006). MGR covers 37,000 ha (Goodman 1990) and KYGR covers 1415 ha (Van Rooyen and Morgan 2007). The reserves are situated 40 km inland, between the Mkhuze River in the north and Phinda Private Game Reserve in the south. The Lebombo Mountains form the western border. The area is characterized by two distinct seasons: a warm and arid winter from April to September, and a hot and humid summer from October to March. Minimum annual average temperature ranges from 12°C to 20°C, and the maximum annual average temperature ranges from 25°C to 34°C. Mean average monthly rainfall ranges between 18 and 38 mm during the dry season and from 61 to 107 mm during the wet season (South African Weather Bureau, Makatini Weather Station).
We selected 20 local study sites at MGR and eight at KYGR in sample areas representing the overall habitat diversity of the reserves (Figure 1, Table 1). To choose the study sites, the maps of the reserves were divided into 1 km×1 km squares. The localizations of the sites were chosen at random within the 1 km×1 km squares falling in the major habitat types.
|Habitat type||Dominant trees||Dominant shrubs||Soil type|
|Acacia woodland||Acacia tortilis, Dichrostachys cinerea, Spirostachys africana, Acacia burkei, Combretum apiculatum, Ziziphus mucronata||Dombeya rotundifolia, Canthium setiflorum, Gymnosporia buxifolia||Low clay content, high calcium content|
|Sand forest||Newtonia hilderbrandtii, Cleistanthus schlechteri, Hymenocardia ulmoides, Pteleopsis myrtifolia, Dialum schlechteri, Croton gratissimus, Stychnos henningsii||Salacia leptoclada, Croton pseudopulchellus, Hymenocardia ulmoides||Sandy, acidic, sillicaceous|
|Combretum molle woodland on red sand||Combretum molle, Acacia burkei, Sclerocarya caffra||Terminalia sericea, Ziziphus mucronata||Ferruginous arenosols|
|Lebombo thicket||Combretum apiculatum, Dichrostachys cinerea, Acacia caffra, Commiphora neglecta, Ziziphus mucronata,||Ozoroa englerophytum, Gymnosporia buxifolia||Rocky|
|Floodplain grassland||Acacia xanthophloea, Ficus sycamorus||Phragmites australis, Echinocloa pyramidalis, Eriochloa spp., Sorgum spp., Cyperus spp.||Alluvial, high clay content|
|Riverine woodland||Acacia xanthophloea, Rauvolfia caffra, Trichilia emetica, Ficus sycamorus||Acacia schweinfurthii, Azima tetracantha, Eupatorium odoratum||Alluvial, high clay content|
|Lebombo wooded grassland||Ziziphus mucronata,||Rocky|
|Ziziphus mucronata bushland||Ziziphus mucronata, Acacia tortilis, Acacia grandicornuta, Acacia nilotica, Acacia senegal, Dichrostachys cinerea||Euclea divinorum, Azima tetracantha, Acacia borleae||High clay content|
|Spirostachys africana woodland||Spirostachys africana, Sideroxylon inerme, Pappea capensis, Ekebergia capensis||Euclea divinorum, Euclea racemosa, Diospyros natalensis, Carissa bispinosa, Catunaregam spinosa||High clay content, poorly drained|
Rodent and shrew sampling
We used both pitfall traps and live traps to capture rodents and shrews during the winter and summer months of 2007, 2008 (MGR), and 2009 (KYGR). At each local study site in MGR, we set up 15 live traps (Scientific Supa Kill CC) and 25 pitfall traps connected by drift fences; in KYGR, we used 15 live traps and four pitfall traps connected by drift fences. Pitfall traps consisted of 20 l buckets that were buried in the ground with the rim of each bucket at ground level. Pitfall traps were 3.5 m apart from each other and arranged at a 120° angle between each line (Figure 2). Live traps were arranged 10 m apart from each other in a 140 m line transect and at least 10 m away from the pitfall traps. Traps were checked every morning, and live traps were re-baited with a mixture of peanut butter and oats (McComb et al. 1991). Study sites were at least 500 m apart from each other to reduce the likelihood that rodents and shrews from one site would disperse to other study sites (Hurst et al. 2013).
We identified rodents in the field by the following external characters: total length, tail length, ear length, and shape of the body (De Graaff 1981, Taylor 1998). We released individuals that could be identified after clipping their dorsal fur to avoid counting them twice. We euthanized shrews and rodent individuals that could not be identified in the field, as well as voucher specimens for each species at each site (these were hosted in the Durban Natural Science Museum, South Africa). Two specimens keyed out as Mus cf. indutus (DM 9265) and Mus cf. neavei (DM 9174) on external and cranial characters (see Lamb et al. 2014 for details) and were conservatively retained as rare species for this study. Their inclusion or exclusion did not materially affect the outcomes of the study.
All handling of living animals was carried out in accordance with guidelines of the American Society of Mammalogists (Gannon et al. 2007) and under a Durban Natural Science Museum collecting permit from Ezemvelo KZN Wildlife.
We measured 17 microhabitat variables at each local study site in winter and summer. We quantified ground cover using the line-intercept method (Mueller-Dombois and Ellenberg 1974). Three 30-m transects were set along the pitfall trap lines. Every 50 cm along each transect, we recorded the following six ground cover variables: percentage bare soil, percentage plant cover, percentage rock cover, percentage shrub cover, percentage log cover, and percentage litter cover. In addition, we measured grass height to obtain a measure of vertical heterogeneity and used these data to classify grass into seven classes: % grass 0–5 cm, % grass 6–10 cm, % grass 11–20 cm, % grass 21–30 cm, % grass 31–40 cm, % grass 41–50 cm, and % grass >50 cm. We calculated the mean densities of trees and shrubs at the end of each pitfall trap line and at the center of the pitfall trap array using the point quarter method (Bonham 1989). Next, we obtained a mean value of canopy cover by measuring at those four points the amount of light coming through the vegetation at ground level using a photoelectric meter. Finally, from the center of the pitfall trap array, we visually assigned a value of slope inclination: 1=flat (∼0°), 2=intermediate (≤15°), and 3=steep (>15°).
We conducted a principal component analysis (PCA; SPSS version 15, LEAD Technologies, Inc., 2006) to reduce the number of variables and remove correlations between the microhabitat variables. Then, we used the principal components (PCs) as new microhabitat parameters in subsequent analyses. Although correspondence analysis can also be used to reduce the number of variables, this is more suitable for nominal data, and we wanted to emphasize the patterns in the actual values of the variables rather than the patterns of relative composition. We analyzed winter and summer data separately to account for the seasonal change in vegetation. We used five principal components in each season based on the percentage cumulative variance they explained.
Using EstimateS (version 8.2, Colwell 2009), we extrapolated expected species richness with two richness estimators, Chao 2 (Chao 1987) and Jackknife 2 (Palmer 1991) in order to assess the completeness of species inventories. These two estimators perform well with small numbers of samples (Colwell and Coddington 1994). By comparing the expected species richness with the observed one, the percentage completeness of a species inventory can be calculated.
Next, we plotted sample-based rarefaction curves using the software EstimateS (version 8.2, Colwell 2009) to compare species richness among study sites. The rarefaction curves standardize sampling effort at different study sites by taking into consideration the heterogeneity of the data (Gotelli and Colwell 2001).
We investigated the relationships between species richness and abundance, and PCs using general linear models (GLM, SPSS version 15, LEAD Technologies, Inc., 2006). Abundances and species richness were square root-transformed to meet the assumptions of normality and equal variances (Kolmogorov-Smirnov test, p>0.2). We used two-way analysis of variance (ANOVA) to determine the effects of season and habitat types, as well as the interaction effect between both factors, on rodent and shrew species richness and abundance. Post-hoc Tukey tests were performed on significant ANOVAs.
Using Ecosim (version 7.71, Gotelli and Entsminger 2012), we quantified species composition patterns with the following co-occurrence indices, in order to test predictions from competition theory: (1) the C-score (Stone and Roberts 1990), which measures the number of checkerboard units of all possible pairs of species (it should be significantly larger than expected by chance); (2) the number of checkerboards (Diamond 1975), which measures the number of species pairs that never coexist at any site (it should be significantly larger than expected by chance); (3) the number of unique species combinations (Pielou and Pielou 1968), which should be lower than expected by chance; and (4) the V-ratio (Schluter 1984), which measures the variability of the number of species per site and tests the niche limitation hypothesis (it should be smaller than expected by chance, Wilson et al. 1987). To test these predictions, observed co-occurrence indices were compared with 1000 expected values generated using the null models identified in MPower (see below). We considered co-occurrence patterns non-random if the observed co-occurrence indices were significantly different from 95% of the expected values. Co-occurrence patterns were analyzed in winter, summer, in both seasons and at different spatial scales. The analyses were performed within and across habitat types of MGR and KYGR, and at the broadest scale encompassing all habitat types of both reserves (MGR+KYGR).
In order to compare observed co-occurrence patterns with patterns expected by chance (Gotelli 2000), Ecosim software can be used to generate nine different null models (named SIM1 to SIM9, respectively; Gotelli and Entsminger 2012), each differing in the way rows and columns of the matrix are randomized. Because these null models may give conflicting conclusions when applied to the same data set, we used the software MPower (Ladau and Ryan 2010) to identify the most suitable SIM1 to SIM9 models for our data set, measured by size (Type 1 error), power (Type 2 error), robustness (dependence of a test’s error rates on assumptions), and bias (a measure of how much more likely the null hypothesis is to be rejected when it is false than when it is true). Species were assumed to have different probabilities of occurring at different sites and species as well as different probabilities of occurring at the same site (Ladau and Ryan 2010).
To test whether rodent and shrew assemblages were hierarchically structured, we quantified nestedness (i.e., whether species in species-poor assemblages are subsets of species in species-rich assemblages) using the nestedness temperature calculator BINMATNEST (Rodríguez-Gironés and Santamaría 2006). We also compared observed nestedness temperature with expected temperatures calculated for 1000 simulations. Nestedness temperatures range from 0 (a set of perfectly nested assemblages) to 100 (a set of completely disordered assemblages). We used Spearman rank correlations to test the respective correlations between the site rank order of study sites in the maximally packed matrix and site isolation, site area, and habitat heterogeneity. Each local study site was encompassed within a continuous unit composed of the two adjacent game reserves, MGR and KYGR, surrounded by disturbed areas. We quantified site isolation using the following indices: distance from the local study site to the nearest and the farthest borders of the unit, distance from the local site to the edge of the habitat patch where the site is found, distance from the site to the nearest patch of the same habitat as the one where it is found, and sum of the pairwise distances between sites (Cullingham et al. 2008). Furthermore, we quantified site area with two indices, namely, size of the habitat patch where the site is found and area of this habitat within the unit. The indices were measured with ArcMap (version 9.3, ESRI Inc., 2008) using the “Measure” tool. We quantified habitat heterogeneity with six indices measured at macrohabitat and microhabitat scales, including number of habitats adjacent to the habitat patch where the site is found and the principal components of the microhabitat variables, among others. Nestedness was analyzed within and across habitat types of each reserve, and across all habitat types of both reserves (MGR+KYGR).
Species richness and abundance
At MGR, a total of 215 rodents representing 14 species and ten genera in two families, Muridae and Nesomyidae, were captured over 102 trapping nights (Table 2). The two most common rodent species caught were Mus minutoides and Mastomys natalensis, representing 59% of all the captures. The least abundant species were Steatomys krebsii, Mus cf. neavei and M. cf. indutus. The species richness estimators indicated that our species inventory was between 64% (Chao 2) and 70% (Jackknife 2) complete. Sample-based rarefaction curves indicated that rodent species richness was the highest in the Acacia woodlands and the lowest in sand forests (Figure 2).
|Acacia woodland (sites 1, 2, 3, 4, 9, 11, 15, 18, 20||Sand forest (sites 6, 7, 10, 12)||Combretum molle woodland on red sand (sites 8, 13, 14)||Lebombo thicket (site 5)||Floodplain grassland (sites 16, 17)||Riverine woodland (site 19)|
|Aethomys ineptus (Thomas and Wroughton 1908)||5||2||7||0||0||0||3||1||4||0||0||0||0||0||0||2||0||2|
|Dendromus melanotis Smith 1834||0||0||0||1||1||2||0||0||0||0||0||0||0||0||0||0||0||0|
|D. mystacalis Heuglin 1863||0||0||0||1||0||1||1||0||1||0||0||0||2||0||2||1||0||1|
|Grammomys dolichurus (Smuts 1832)||0||0||0||0||1||1||0||0||0||0||0||0||0||0||0||0||0||0|
|Lemniscomys rosalia (Thomas 1904)||4||1||5||2||3||5||0||0||0||4||1||5||0||0||0||0||0||0|
|Mastomys natalensis (Smith 1834)||35||2||37||7||0||7||1||0||1||12||0||12||2||0||2||1||0||1|
|Mus cf. indutus (Thomas 1910)||1||0||1||0||0||0||0||0||0||0||0||0||0||0||0||0||0||0|
|M. minutoides Smith 1834||16||10||26||11||12||23||6||2||8||2||0||2||6||0||6||2||0||2|
|M. cf. neavei (Thomas 1910)||0||0||0||1||0||1||0||0||0||0||0||0||0||0||0||0||0||0|
|Saccostomus campestris Peters 1846||4||1||5||0||0||0||2||0||2||2||0||2||2||0||2||0||0||0|
|Steatomys krebsii Peters 1852||1||0||1||0||0||0||0||0||0||0||0||0||0||0||0||0||0||0|
|S. pratensis Peters 1846||2||5||7||0||0||0||0||0||0||0||0||0||0||0||0||0||0||0|
|Gerbilliscus leucogaster (Peters 1852)||3||2||5||6||2||8||6||1||7||0||0||0||1||0||1||0||0||0|
|Thallomys paedulcus (Sundevall 1846)||3||3||6||2||0||2||0||0||0||0||0||0||1||0||1||0||0||0|
|Crocidura hirta Peters 1852||8||18||26||1||2||3||1||2||3||2||1||3||2||0||2||0||0||0|
|C. fuscomurina (Heuglin 1865)||4||7||11||0||1||1||2||1||3||1||0||1||2||0||2||11||0||11|
|C. silacea Thomas 1895||3||2||5||3||3||6||0||1||1||0||0||0||0||0||0||0||0||0|
|Suncus lixus (Thomas 1898)||5||2||7||3||2||5||3||0||3||0||0||0||2||0||2||1||0||1|
At MGR, we captured 96 shrews representing four species and two genera from the family Soricidae over 102 trapping nights (Table 2). The two most commonly species caught were Crocidura fuscomurina and C. hirta, representing 73% of all captures. Crocidura silacea was the least abundant species caught. Both estimators indicated that the species inventory of shrews was 100% complete. Species richness at a local scale was the highest in the Acacia woodlands and the lowest in sand forests (Figure 3
At KYGR, a total of 63 rodents representing six species and six genera belonging to two families, Muridae and Nesomyidae, were captured over 20 trapping nights (Table 3). The two most common rodent species caught were Mus minutoides and Aethomys ineptus, representing 70% of all the captures. The least abundant species were Lemniscomys rosalia and Mastomys natalensis. The species richness estimators indicated that our rodent species inventory was between 83% (Jackknife 2) and 100% (Chao 2) complete. Meanwhile, sample-based rarefaction curves indicated that species richness was the highest in Lebombo wooded grasslands and the lowest in riverine woodlands (Figure 3).
|Lebombo wooded grassland (sites 1, 2)||Ziziphus mucronata bushland (sites 3, 6)||Riverine woodland (sites 4, 7)||Spirostachys africana woodland (sites 5, 8)|
|Suncus infinitesimus (Heller 1912)||5||0||5||0||0||0||0||0||0||0||0||0|
At KYGR we captured 21 shrews representing four species and two genera from the family Soricidae over 20 trapping nights (Table 3), these included Crocidura hirta (six individuals), Suncus lixus (six individuals), S. infinitesimus (five individuals), and C. silacea (four individuals). Both estimators indicated that the species inventory of shrews was 100% complete. Suncus infinitesimus was collected at KYGR but not at MGR, while Crocidura fuscomurina was collected at MGR but not at KYGR. Species richness at a local scale was the highest in Lebombo wooded grasslands and the lowest in Ziziphus mucronata bushland (Figure 3).
At identical sampling efforts (i.e., for the same number of individuals caught), rodent species richness was higher at MGR than at KYGR, while shrew species richness was higher at KYGR than at MGR (Figure 4). There were also significant differences, but only in rodent species richness and abundance between seasons (F1,8=7.05, p<0.05; F1,8=7.80, p<0.05, respectively). Post-hoc Tukey tests revealed that rodent species richness and abundance were the highest in winter.
Correlation with microhabitat characteristics
The PCA of the 17 microhabitat variables extracted five principal components that accounted for 79.48% of the total variance in winter and 77.11% in summer (Table 4). We interpreted the PCs as follows. In winter, PC1 was a measure of differences in the vertical height of grass and % plants: local study sites with a high % grass height 31–40 cm and 41–50 cm, % plant cover loaded high on the axis and sites with a high % grass height 0–5 cm loaded low on the axis. PC2 was a measure of differences in tree density and % litter: local study sites with a high density of trees and high % litter loaded high on the axis. PC3 was a measure of differences in canopy cover and % grass height >50 cm: local study sites with a high canopy cover loaded high on the axis and sites with a high % grass height >50 cm loaded low. PC4 was a measure of differences in the percentage of rocks: local study sites with a high percentage of rocks loaded high on the axis. Finally, PC5 was a measure of differences in the percentage of bare soil: local study sites with a high percentage of bare soil loaded low.
|% bare soil||-0.600||-0.190||0.412||0.131||-0.419||-0.463||-0.515||0.495||0.123||0.292|
|Grass 0–5 cm||-0.899||0.149||-0.006||0.041||-0.28||-0.801||-0.192||0.125||-0.069||0.394|
|Grass 6–10 cm||0.235||0.667||0.260||0.007||0.361||-0.422||-0.178||-0.325||-0.544||-0.226|
|Grass 11–20 cm||0.406||0.438||0.641||0.050||0.335||-0.075||0.007||0.452||-0.425||-0.565|
|Grass 21–30 cm||0.723||-0.024||0.522||0.072||0.049||0.593||0.308||0.487||0.229||-0.323|
|Grass 31–40 cm||0.808||-0.245||0.050||0.211||-0.112||0.798||0.157||0.383||0.309||0.015|
|Grass 41–50 cm||0.834||-0.264||-0.229||0.086||-0.062||0.873||0.182||-0.125||0.177||0.229|
|Grass >50 cm||0.214||-0.526||-0.645||-0.243||0.146||0.312||0.028||-0.681||0.425||0.233|
|Total variance explained (%)||30.37||20.85||13.35||7.54||7.36||33.08||13.68||12.24||10.10||7.99|
|Cumulative variance (%)||30.37||51.22||64.57||72.12||79.48||33.08||46.76||59.00||69.10||77.09|
In summer, PC1 was a measure of differences in the vertical height of grass: local study sites with a high % grass height 31–40 cm and 41–50 cm loaded high on the axis, and those with a high % grass height 0–5 cm loaded low on the axis. PC2 was a measure of differences in % logs and % litter: local study sites with high % logs and high % litter loaded high on the axis. PC3 was a measure of differences in canopy cover and % grass height >50 cm: local study sites with a high canopy cover loaded high on the axis and sites with a high % grass height >50 cm loaded low. PC4 was a measure of differences in the percentage of shrubs and % grass height 6–10 cm: local study sites with a high % of shrubs loaded high on the axis and sites with a high % grass height 6–10 cm loaded low. Finally, PC5 was a measure of differences in tree density and % grass height 11–20 cm: local study sites with a high tree density loaded high on the axis and sites with a high % grass height 11–20 cm loaded low.
Only in winter, rodent species richness was significantly correlated with PC1 (r27=0.4, p<0.05) and PC4 (r27=0.6, p<0.05), and rodent abundance was significantly correlated with PC1 (r27=0.6, p<0.05). Similarly, only in winter, shrew species richness was significantly correlated with PC1 (r27=0.3, p<0.05), PC2 (r27=0.4, p<0.05), and PC4 (r27=0.5, p<0.05).
Non-random co-occurrence patterns
The influence of competition on rodent and shrew assemblages was tested with four co-occurrence indices and seven algorithms (Table 5). Based on the assessments of the quality of simulations, co-occurrence patterns within habitat types could only be tested in the Acacia woodlands and not in the shrew assemblages at KYGR.
|Index||Randomizing algorithm||Obs index||Sim index||p-Value||Index||Randomizing algorithm||Obs index||Sim index||p-Value||Index||Randomizing algorithm||Obs index||Sim index||Index|
|Mkhuze Game Reserve (MGR)|
|Kube Yini Game Reserve (KYGR)|
See Gotelli and Entsminger (2012) for details of randomizing algorithms See text for details of indices; CB, number of checkerboards; SC, number of unique species combinations. The p-values in bold indicate significant patterns consistent with competition predictions.
In winter, we found non-random patterns consistent with competition theory in rodent assemblages at MGR and MGR+KYGR with SIM1 in combination with the number of checkerboards (Table 5). This algorithm allows the number of species at a site to vary, with the constraint that all sites have the same average number of species, and occurrence frequencies of each species vary with the same probability. In both seasons, we found non-random patterns consistent with competition theory in rodent assemblages at MGR and MGR+KYGR with SIM1 in combination with the number of checkerboards. We did not find patterns of competition in summer.
Rodent assemblages were significantly nested at MGR in winter, summer, and both seasons combined; they were significantly nested at KYGR in both seasons combined; and they were significantly nested at MGR + KYGR in summer and both seasons combined (Table 6). However, results from the correlation analyses were only significant at the broadest scale MGR+KYGR with both seasons combined. The site rank order in the maximally packed matrix was significantly correlated with the distance from the local study site to the nearest (r27=0.4, p<0.05) and the farthest (r27=0.4, p<0.05) borders of the unit formed by the two reserves, the distance from the local study site to the nearest patch of the same habitat as the one where the local study site can be found (r27=0.5, p<0.05), the sum of the pairwise distances between sites (r27=0.4, p<0.05), the size of the habitat patch where the local study site can be found (r27=0.4, p<0.05), the size of this habitat in the unit formed by the two reserves (r27=0.4, p<0.05), the PC2 (r27=0.5, p<0.05), and the PC4 (r27=0.4, p<0.05) of the microhabitat variables measured in winter.
|Combretum molle woodland||Rodents||–||–||–||–||0.34||17.3|
|Mkhuze Game Reserve (MGR)||Rodents||0.02||18.2||0.04||10.3||0.004||17.1|
|Kube Yini Game Reserve (KYGR)||Rodents||0.1||3.8||0.4||18.7||0.03||7.1|
T, observed temperature of the nested matrix. Values in bold are significantly nested at p<0.05.
We found non-random patterns consistent with competition theory in shrew assemblages in winter at MGR with SIM4 in combination with C-score, and in both seasons at MGR+KYGR with SIM1 in combination with the number of checkerboards (Table 5). In SIM4, the species occurrence totals were maintained, and the probabilities of occurrence in sites were proportional to the observed species richness per site.
Shrew assemblages were only significantly nested across habitats at MGR+KYGR when both seasons were combined (Table 6). The site rank order in the maximally packed matrix was only significantly correlated with the PC3 of the microhabitat variables measured in summer at the broadest scale MGR+KYGR (r27=0.5, p<0.05).
A total of 14 rodent and five shrew species were captured at both reserves, including five species, Dendromus melanotis, Steatomys pratensis, S. krebsii, Mus cf. indutus and M. cf. neavei that have not previously been recorded from MGR (Taylor 1998; Taylor et al. 2007). All the species captured at KYGR are new records for the reserve. The latter two species are new records for the Province of KwaZulu-Natal, but their taxonomic status, particularly that of M. cf. neavei, must be verified with multiple lines of morphological and molecular evidence (Lamb et al. 2014). Two rodent and two shrew species (i.e., Mastomys natalensis, Mus minutoides, Crocidura hirta and C. fuscomurina, respectively) represented most of the captures. These species are widely distributed in southern Africa and have a broad habitat tolerance (Monadjem 1997, 1999b). On the one hand, based on two different species richness estimators, our inventory of the rodent species pool at MGR was between 64% and 70% complete, suggesting that rodent species richness may be higher. At KYGR, on the other hand, the rodent species pool was between 83% and 100% complete. The richness estimators indicated that the shrew inventories at both MGR and KYGR were 100% complete. By comparison, small mammal species richness at habitat and reserves scales were lower in other studies in the savannah biome (e.g., Linzey and Kesner 1997, Caro 1999, 2001), including the neighboring Phinda Private Game Reserve (Rautenbach et al. 2014).
Rodent species richness and abundance were higher in the dry season than in the wet season, despite savannah plant diversity being higher in the wet season. Sites with high grass as well as sufficient ground and canopy cover harbor a greater number of species because they provide more food (Monadjem and Perrin 2003) and better protection against predators (Kelt et al. 2004, Kotler et al. 1991) than open sites. However, similar seasonal patterns of rodent species composition (i.e., increased species richness in the dry season) have been recorded in southern Africa (Mahlaba and Perrin 2003, Monadjem and Perrin 2003, Schradin and Pillay 2006). One reason may be a delayed response in the temporal availability of resources (Hansen et al. 1999, Hernández et al. 2005). Additionally, the higher food availability may have rendered the bait in traps less attractive to the rodents during the wet season than during the dry season when food abundance is low (Monadjem 1999a, dos Santos-Filho et al. 2006). It is notable that in the dry season, but not in the wet season, rodent and shrew species richness and rodent abundance were significantly positive correlated with microhabitat features, such as grass height, tree density, and ground cover.
We found significantly nested patterns in rodent assemblages at the reserve spatial scale (both for individual reserves and combined), but not at the habitat (Acacia woodlands) spatial scale. We also found significant nested patterns in shrew assemblages at the reserve spatial scale (when both reserves were combined). For rodent assemblages, there were strong correlations between site area and nestedness and site isolation, thereby suggesting the influence of immigration and extinction on species composition patterns (Patterson and Atmar 1986). Conversely, no correlation was found between nestedness and site isolation or site area for shrew assemblages. This suggested that large scale biogeographic processes may be important in structuring rodent assemblages, whereas processes operating at smaller scales may be important for nestedness in shrew assemblages.
Nested hierarchies among species may also be produced by a pattern of included niches, such that species with narrow tolerances for environmental conditions represent subsets of species with broad tolerances for environmental conditions (Abu Baker and Patterson 2011). For example, differential tolerances to elevations and climate conditions probably produced the nested pattern observed in North American rodent assemblages (Kelt et al. 1999). In this study, the positive correlations between rodent nestedness and the percentage of rocks, the percentage of litter and tree density as well as that between shrew nestedness and canopy cover and the percentage of tall grass all suggested that habitat filtering operating at a microhabitat scale produced the nested patterns in rodent and shrew species composition. Furthermore, species with specific requirements for these microhabitat features represent subsets of species with no particular preferences for these features.
We found little support for predictions of competition hypotheses, which asserts that in competitively structured assemblages, there should be fewer species combinations and more checkerboard species pairs, and the variance of species richness among sites should have been smaller than expected by chance (Gotelli and Entsminger 2012). Only four out of 54 null model simulations for the rodents and two out of 44 simulations for the shrews displayed significantly more checkerboard species pairs than expected by chance. Similarly, no evidence for non-random species composition patterns consistent with competition hypotheses was found in rodent and shrew assemblages at the neighboring Phinda Private Game Reserve (Rautenbach et al. 2014). Conversely, non-random patterns of rodent and shrew species co-occurrence consistent with competition theory were detected in temperate and desert regions (Brown et al. 2000, Kelt et al. 1995, 1999). Perhaps species inhabiting these regions are more likely to compete for limited resources than in the savannah biome where resource availability is higher.
To conclude, rodent and shrew diversity showed seasonal fluctuations and were significantly correlated with microhabitat features. In addition, rodent and shrew assemblages displayed significant nested patterns, suggesting that abiotic processes, specifically habitat filtering at a local scale, and immigration and extinction at a regional scale, are driving assemblage structure while competition plays a minor role, if any. Therefore, future studies should consider the relative influence of other biotic processes (e.g., predation) and quantify other niche dimensions (e.g., diet) using stable isotopes, for example (Symes et al. 2013), in order to further unravel potential niche partitioning among small mammal species in the savannah biome.
We thank the eco-volunteers, students from UKZN, and the management of Mkhuze Game Reserve for the logistic support and assistance they have provided in the field. We are grateful to Ian Macdonald for providing us accommodation and logistic support at Kube Yini Game Reserve. MCS acknowledges financial support from the University of KwaZulu-Natal (competitive research grant). PJT acknowledges funding from the National Research Foundation. We are also grateful to the two anonymous referees for their helpful comments and suggestions, which helped improve our manuscript.
Abu Baker, M. and B.D. Patterson. 2011. Patterns in the local assembly of Egyptian rodent faunas: Co-occurrence and nestedness. J. Arid Environ. 75: 14–19.10.1016/j.jaridenv.2010.08.005Search in Google Scholar
Avenant, N.L. and P. Cavallini. 2007. Correlating rodent community structure with ecological integrity, Tussen-die-Riviere Nature Reserve, Free State province, South Africa. Integr. Zool. 2: 212–219.10.1111/j.1749-4877.2007.00064.xSearch in Google Scholar
Bonham, C.D. 1989. Measurements for terrestrial vegetation. Wiley, New York.Search in Google Scholar
Caro, T.M. 1999. Abundance and distribution of mammals in Katavi National Park, Tanzania. J. App. Ecol. 36: 205–217.Search in Google Scholar
Clark, J.E., E.C. Hellgren, J.L. Parsons, E.E. Jorgensen, D.M. Engle and D.M. Leslie Jr. 2005. Nitrogen outputs from faecal and urine deposition of small mammals: implications for nitrogen cycling. Oecologia 144: 447–455.10.1007/s00442-005-0004-9Search in Google Scholar PubMed
Colwell, R.K. 2009. EstimateS: Statistical estimation of species richness and shared species from samples. Version 8.2. User’s Guide and application published at: http://purl.oclc.org/estimates.Search in Google Scholar
Cullingham, C.I., B.A. Pond, C.J. Kyle, E.E. Rees, R.C. Rosatte and B.N. White. 2008. Combining direct and indirect genetic methods to estimate dispersal for informing wildlife disease management decisions. Mol. Ecol. 17: 4874–4886.10.1111/j.1365-294X.2008.03956.xSearch in Google Scholar PubMed
De Graaff, G. 1981. The rodents of southern Africa. Butterworth, Durban, South Africa.Search in Google Scholar
Diamond, J.M. 1975. Assembly of species communities. In: (M.L. Cody and J.M. Diamond, eds.) Ecology and evolution of communities. Cambridge University Press, Cambridge. pp. 342–444.Search in Google Scholar
dos Santos-Filho, M., D.J. da Silva and T.M. Sanaiotti. 2006. Efficiency of four trap types in sampling small mammals in forest fragments, Mato Grosso, Brazil. Mastozool. Neotr. 13: 217–225.Search in Google Scholar
Du Toit, R., G. Hinde, and W. Taylor. 2001. Africa’s big five. Struik Publishers, Cape Town.Search in Google Scholar
Fischer, J. and D.B. Lindenmayer. 2005. Nestedness in fragmented landscapes: a case study on birds, arboreal marsupials and lizards. J. Biogeogr. 32:1737–1750.10.1111/j.1365-2699.2005.01319.xSearch in Google Scholar
Gannon, W.L., R.S. Sikes, and the Animal Care and Use Committee of the American Society of Mammalogists. 2007. Guidelines of the American Society of Mammalogists for the use of wild mammals in research. J. Mammal. 88: 809–823.10.1644/06-MAMM-F-185R1.1Search in Google Scholar
Goodman, P.S. 1990. Soil, vegetation and large herbivores relations in Mkuzi Game Reserve, Natal, PhD thesis. University of the Witwatersrand, Johannesburg.Search in Google Scholar
Gotelli, N.J. and G.R. Graves. 1996. Null models in ecology. Smithsonian Institution Press, Washington DC.Search in Google Scholar
Gotelli, N.J. and R.K. Colwell. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett. 4: 379–391.10.1046/j.1461-0248.2001.00230.xSearch in Google Scholar
Gotelli, N.J. and G.L. Entsminger. 2012. EcoSim: null models software for ecology. Version 7.71. Acquired Intelligence Inc., Kesey-Bear. http://homepages.together.net/∼gentsmin/ecosim.htm.Search in Google Scholar
Hansen, T.F., N.C. Stenseth and H. Henttonen. 1999. Multiannual vole cycles and population regulation during long winters: an analysis of seasonal density dependence. Am. Nat. 154: 129–139.10.1086/303229Search in Google Scholar
Hernández, L., A.G. Romero, J.W. Lanundré, D. Lightfoot, E. Aragón and J. López Portillo. 2005. Changes in rodent community structure in the Chihuahuan Desert Mexico: comparisons between two habitats. J. Arid Environ. 60: 239–257.10.1016/j.jaridenv.2004.03.013Search in Google Scholar
Horner-Devine, M.C., J.M. Silver, M.A. Leibold, B.J.M. Bohannan, R.K. Colwell, J.A. Fuhrman, J.L. Green, C.L. Kuske, J.B.H. Martiny, G. Muyzer, L. Øvreås, A.-L. Reysenbach, and V.H. Smith. 2007. A comparison of taxon co-occurrence patterns for macro- and microoganisms. Ecology 88: 1345–1353.10.1890/06-0286Search in Google Scholar
Hurst, Z.M., R.A. McCleery, B.A. Collier, R.J. Fletcher, N.J. Silvy, P.J. Taylor and A. Monadjem. 2013. Dynamic edge effects in small mammal communities across a conservation-agricultural interface in Swaziland. PLoS One 8: e74520.10.1371/journal.pone.0074520Search in Google Scholar
Kelt, D.A., M.L. Taper and P.L. Meserve. 1995. Assessing the impact of competition on community assembly: a case study using small mammals. Ecology 76: 1283–1296.10.2307/1940935Search in Google Scholar
Kelt, D.A., K.A. Rogovin, G. Shenbrot and J.H. Brown 1999. Patterns in the structure of Asian and North American desert small mammal communities. J. Biogeogr. 26: 825–841.10.1046/j.1365-2699.1999.00325.xSearch in Google Scholar
Kelt, D.A., P.L. Meserve, L.K. Nabors, M.L. Forister and J.R. Gutiérrez. 2004. Foraging ecology of small mammals in semiarid Chile: the interplay of biotic and abiotic effects. Ecology 85: 383–397.10.1890/02-4065Search in Google Scholar
Lamb, J., S. Downs, S. Eiseb and P.J. Taylor. 2014. Increased geographic sampling reveals considerable new genetic diversity in the morphologically conservative African Pygmy Mice (Genus Mus; Subgenus Nannomys). Mammal. Biol. 79: 24–35.10.1016/j.mambio.2013.08.010Search in Google Scholar
Lawton, J.H. 2000. Community ecology in a changing world. Ecology Institute, Oldendorf/Luhe.Search in Google Scholar
Linzey, A.V. and M.H. Kesner. 1997. Small mammals of a woodland-savanna ecosystem in Zimbabwe. I. Density and habitat occupancy patterns. J. Zool. 243: 137–152.10.1111/j.1469-7998.1997.tb05760.xSearch in Google Scholar
Mahlaba, T.A.M. and M. Perrin. 2003. Population dynamics of small mammals at Mlawula, Swaziland. Afr. J. Ecol. 41: 317–323.Search in Google Scholar
McCay, T.S., M.J. Lovallo, W.M. Ford, and M.A. Menzel. 2004. Assembly rules for functional groups of North American shrews: effects of geographic range and habitat partitioning. Oikos 107: 141–147.10.1111/j.0030-1299.2004.13244.xSearch in Google Scholar
McComb, W.C., G.A. Robert, and K. McGarigal. 1991. Differential vulnerability of small mammals and amphibians to two trap types and two trap baits in Pacific Northwest forests. Northwest Sci. 65: 109–115.Search in Google Scholar
Monadjem, A. 1999a. Geographic distribution patterns of small mammals in Swaziland in relation to abiotic factors and human land-use activity. Biodivers. Conserv. 8: 223–237.Search in Google Scholar
Monadjem, A. 1999b. Population dynamics of Mus minutoides and Steatomys pratensis (Muridae: Rodentia) in a subtropical grassland in Swaziland. Afr. J. Zool. 37: 202–210.10.1046/j.1365-2028.1999.00169.xSearch in Google Scholar
Monadjem, A. and M. Perrin. 2003. Population fluctuations and community structure of small mammals in a Swaziland grassland over a three-year period. Afr. Zool. 38: 127–137.10.1080/15627020.2003.11657200Search in Google Scholar
Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and methods of vegetation ecology. Wiley, New York.Search in Google Scholar
Patterson, B.D. and W. Atmar. 1986. Nested subsets and the structure of insular mammalian faunas and archipelagos. Biol. J. Linn. Soc. 28: 65–82.10.1111/j.1095-8312.1986.tb01749.xSearch in Google Scholar
Pielou, D.P. and E.C. Pielou. 1968. Associations among species of infrequent occurrence: the insect and spider fauna of Polyporus betulinus (Bulliard) fries. J. Theor. Biol. 21: 202–216.10.1016/0022-5193(68)90070-2Search in Google Scholar
Rautenbach, A., T. Dickerson and M.C. Schoeman. 2014. Diversity of rodent and shrew assemblages in different vegetation types of the savannah biome in South Africa: no evidence for nested subsets or competition. Afr. J. Ecol. 52: 30–40.10.1111/aje.12081Search in Google Scholar
Rodríguez-Gironés, M.A. and L. Santamaría. 2006. A new algorithm to calculate the nestedness temperature of presence–absence matrices. J. Biogeogr. 33: 924–935.10.1111/j.1365-2699.2006.01444.xSearch in Google Scholar
Rutherford, M.C., L. Mucina, M.C. Lötter, G.J. Bredenkamp, J.H.L. Smit, C.R. Scott Shaw, D.B. Hoare, P.S. Goodman, H. Bezuidenhout, L. Scott, F. Ellis, L.W. Powrie, F. Siebert, T.H. Mostert, B.J. Henning, C.E. Venter, K.G.T. Camp, S.J. Siebert, W.S. Matthews, J.E. Burrows, L. Dobson, N. van Rooyen, E. Schmidt, P.J.D. Winter, P.J. du Preez, R.A. Ward, S. Williamson and P.J.H. Hurter. 2006. Savanna biome. In: (L. Mucina, M.C. Rutherford, eds.) The vegetation of South Africa, Lesotho and Swaziland, vol 19. South African National Biodiversity Institute, Pretoria. pp. 439–539.Search in Google Scholar
Schluter, D. 1984. A variance test for detecting species associations, with some example applications. Ecology 65: 998–1005.Search in Google Scholar
Schoeman, M.C. and D.S. Jacobs. 2008. The relative influence of competition and prey defenses on the phenotypic structure of insectivorous bat ensembles in southern Africa. PLoS One 3: e3715.10.1371/journal.pone.0003715Search in Google Scholar PubMed PubMed Central
Schradin, C. and N. Pillay. 2006. Female striped mice (Rhabdomys pumilio) change their home ranges in response to seasonal variation in food availability. Behav. Ecol. 17: 452–458.10.1093/beheco/arj047Search in Google Scholar
Symes, C.T, J.W. Wilson, S.M. Woodborne, Z.S. Shaikh and M. Scantlebury. 2013. Resource partitioning of sympatric small mammals in an African forest-grassland vegetation mosaic. Austral. Ecol. 38: 721–729.10.1111/aec.12020Search in Google Scholar
Taylor, P.J. 1998. The smaller mammals of KwaZulu-Natal. University of Natal Press, Pietermaritzburg, South Africa.Search in Google Scholar
Taylor, P.J., A. Rautenbach, M.C. Schoeman and X. Combrink. 2007. A winter survey of the smaller mammals of the uMkhuze Section of the iSimangaliso Wetland Park, KwaZulu- Natal Province, South Africa. Operation Wallacea Ltd., Lincolnshire. www.opwall.com.Search in Google Scholar
Van Rooyen, N. and S. Morgan. 2007. Habitats of Phinda and Mkhuze Game Reserves. Internal report, Ekotrust CC, Lynnwood, South Africa.Search in Google Scholar
Weiher, E. and P.A. Keddy. 2001. Ecological assembly rules: perspectives, advances, retreats. Cambridge University Press, London.Search in Google Scholar
©2015 by De Gruyter