Results
A total of 61 008 ant specimens were collected, comprising 74 species in 25 genera, and five subfamilies. Myrmicinae (47 species and eight genera) was the most diverse and abundant subfamily, followed by Ponerinae (13 species and eight genera), and Formicinae (15 species and six genera). The most specious genera were Tetramorium (22 species), Monomorium and Pheidole each had six species, while Camponotus , Lepisiota , and Pheidole had five species each. Three quarters of all the ant individuals caught belonged to one species,Anoplolepis custodiens . This species, and those in the genusPheidole , accounted for 90% of all the ants.
The best model for species richness included season and habitat (closed/open), with richness being significantly higher in the open habitat and during the wet season (Table 1). The second-best model included soil PC1, with very weak evidence for an increase in richness in more sandy soils (estimate = 0.014 ± 0.015). However, this trend was reversed when considering Hill no.1, with strong evidence that there was significantly less Shannon diversity in the open habitats. Similar to richness, Soil-pc1 was included in the second model, with weaker evidence for a positive response to sandy soil (estimate = 0.026 ± 0.05). The negative impact of open habitats became even more pronounced when Simpsons dominance was modelled. Soil-pc1 was also included in the second-best model. All three models explained around 22% of the variation observed, while at least half of this variation was linked to the random factor, i.e. site-specific characteristics (Table 1).
Site-specific characteristics not measured in this study explained the largest amount of variation in ant assemblage structure (Table 2, Figure 2). This impact of sites on assemblages was exemplified by the distinct ant assemblages comprised of Odontomachus troglodytes ,Tetramorium UKZN_14 (simillimum gp.), andLepisiota UKZN_02 (capensis gp.), associated with the sandy soils of Mansinya (Figure 2 & Figure S1).
Habitat followed seasons in explaining the most variation in species composition, particularly when considering conditional variation explained (Figure 2). None of the structural and soil variables explained significant amounts of variation beyond that explained by the previous three variables, although the first soil principle component (increasing sandy soil, see Figure S2), explained the third largest amount of marginal variation (Table 2). Similarly, no taxa were associated with any of the gradients in habitat structure, while eight indicator taxa were associated with the sand-clay gradient in the study site (Figure 3). A. custodiens notably preferred the clay soils across open and closed habitats. The activities of Odontomachus troglodytes increased in more sandy soil in both open and closed habitats, while Anoplolepis custodiens and TetramoriumUKZN_22 (gabonense gp.) had their highest activities in clay soil. Myrmicaria natalensis , Bothroponera soror andMonomorium junodi activities peaked in loamy soils.
Five species were distributed across all sites, and mostly sampled in both the closed and open habitat types. Three of these, PheidoleUKZN_01 (megacephala gp.), Mesoponera caffraria andMonomorium UKZN_03 were associated with open habitat types, while two species, Pheidole crassinoda ruspolii andTetramorium UKZN_18 (squaminode gp.) were associated with closed habitat (Figure S3).