Discussion
Although site-specific characteristics and seasonality dominated ant assemblage structure, habitat type (open vs closed) had important impacts on ant richness and composition. Open habitat had significantly more species, but were less even than closed habitats; and, although ants responded to gradients in soil characteristics, responses to habitat structure changes were not gradual (Figure S4). This is evidenced by the relatively larger role of categorial classification (open versus closed) in explaining ant diversity. Responses to thickening are, therefore, predicted to be abrupt. This threshold response can partly be explained by the presence of the numerical dominant A. custodiens in open habitats of some sites that were mainly on clayey soils and Odontomachus traglodytes which is a specialised as well as aggressive predator in sandy soils.
Andersen (2019) reasoned that the amount of vegetation cover is an important driver of both flora and fauna communities. However in contrast to our study, Nooten, Schultheiss, Rowe, Facey, and Cook (2019) reported fewer ant species in open habitat. Notably their study aite was in highly altered urban gardens of Greater Sydney Region of south-east Australia. Pacheco and Vasconcelos (2012) also reported fewer species in the least structurally complex habitats of the Brazilian Cerrado. In contrast, Lassau and Hochuli (2004), Yusah and Foster (2016) and Ahuatzin et al. (2019) reported a negative response of ant species richness and diversity with habitat complexity. Similarly, Hethcoat et al. (2019) reported more open and warmer pasturelands supporting more ants than the compared primary or secondary forest.
Overall, habitat explained the third largest amount of conditional variation in the ant assemblages of this study. Similarly, across the western Soutpansberg mountains, vegetation types explained a significant amount of variation in ant assemblages (Munyai & Foord, 2012) . This is not surprising, given that habitat structure had been thought to determine the movements of ants as they are usually impeded by grass stems or woody twigs (Nooten et al., 2019). In the Amazon basin, Guilherme et al. (2019) reported the effects of habitat complexity on ant species foraging strategies and behaviour.
Not surprisingly, ant assemblages in the two habitats were significantly different. It is, however, expected that natural characteristics in a site, for example the amount of sunlight intensity, plant cover and overall variation in habitat structure must have a clear effect on the ant assemblage richness and composition (Campos, Soares, Martins, & Ribeiro, 2006). This is largely because of the effect of ground surface rugosity, which affects foraging activities, for example ant running speed (Greve, Blaha, Teuber, Rothmaier, & Feldhaar, 2019). In this study, closed habitat is largely structural more complex. In total, 24 % of the species were only recorded in the open habitat, compared to the 8 % restricted to closed habitat. This, once again, emphasises the important role of habitat openness, as detailed in Andersen (2019), and references there in, as a crucial driver of ant assemblage composition. However, this larger number of species in the open habitat could also be explained by the species-area effect. Currently, open Savannas dominate the Hluhluwe–iMfolozi Park is largely dominated by open savanna vegetation type (Charles-Dominique, Staver, Midgley, & Bond, 2015), covering >75% of the landscape. This in itself would explain the larger number of species observed in the open habitat.
One site, Mansinya, had a distinct ant assemblage and therefore highlighting on the importance of site-specific characteristics, particularly that related to soil characteristics. The Mansinya site is on sandy soil with specific species associated with this site (Figure S2). Similarly, sites on their own were the most influential predictors of ant assemblage structure, as suggested by both conditional and marginal effects (Table 1). Whether these differences are related to historical factors, spatial autocorrelation or variables that we did not measure is difficult to say and beyond the scope of this study. In general, Radnan, Gibb, and Eldridge (2018) noted that ants tend to be more responsive to small-scale alteration on the soil surface than overall changes in vegetation community composition. The influence of site-specific characteristics, namely, vegetational complexity, litter, and soil chemical properties were reported in Parui, Chatterjee, and Basu (2015) as the driver of variation in ant assemblage composition in the forested habitats of Eastern India.
The activities of the hyperabundant (72 % of the total abundance)Anoplolepis custodiens , increased in site with clay soils (Figure 3). In contrast, activities of Odontomachus troglodytes peaked in sandy soils. Three other species (Myrmicaria natalensis ,Bothroponera soror and Monomorium junodi ), were more active in loamy soils. Along an alluvial fan of the Tehuacán Valley of central Mexico, Rios-Casanova, Valiente-Banuet, and Rico-Gray (2006) reported high ant abundance in sandy soils. The latter study suggested that sandy soils and habitat complexity could have favoured the abundance of the dominant ants, which in turn could have also affected overall ant diversity patterns in their study. Bownes, Moore, and Villet (2014) also observed very large number of Anoplolepis custodiensin in citrus orchards of the Eastern Cape, South Africa while A. custodiens has been observed to nest in open and well insulated soils of the Karoo (Dean, 1992).
The high abundance, dominant behaviour, and impact of A. custodiens on ant diversity have been documented in several southern African studies (Mauda, Joseph, Seymour, Munyai, & Foord, 2018; Parr, 2008; Samways, 1983). Although this species is native to the region, it is well known to prey on a variety of other invertebrates (Keiser, Wright, & Pruitt, 2015), and also small livestock (Prins, Robertson, & Prins, 1990), it is also a major pest in agriculture landscapes (Addison & Samways, 2006; Addison & Samways, 2000). A. custodiens is also well known for tending honeydew producing pest insect species, and being nectar collectors, hence, regarded as one of the serious pests in the agricultural systems (Dean, 1992).
A. custodiens respond positively to disturbance (Addison & Samways, 2000; Mauda et al., 2018). Coupled with the formation of supercolonies in open shallow soils in particular (TC Munyai Personal observation), suggests that manual removal of trees and shrubs should not be done indiscriminately. The fact that this species can completely dominate open habitat (Addison & Samways, 2006), clearing of sites could facilitate invasion with negative impacts not only on ant diversity, but also ecosystem services. Particularly since A. custodiens is a polygynous (Samways, 1990), generalist predator of other invertebrates (Dean, 1992), and small vertebrates (Mauda et al., 2018; Prins et al., 1990), displacing other ant species (Mitchell, 2000), and control ant assemblage structure in the South African savanna ( Parr, 2008).
A. custodiens is particularly active in clay soils, and clearing of woody species in the park need to be context specific. Exploring the response of A. custodiens to various woody species control measures, be it manual removal of trees and shrubs, using fire or browsing mammals as a management tool, should provide a better understanding of possible impacts. Generally though, we consider this a cautionary note around predicting local impacts based on broad-scale global patterns.