Spatial mosaics of small animal communities across habitat and vegetation cover in semiarid Australia

Introduction

Arid lands are characterized by ecosystem processes that create substantial large-scale spatial and temporal variation in the landscape and biota. Variation is driven largely by rainfall events that are often unpredictable across space, time, and intensity in arid lands (Chesson et al. 2004; Loik et al. 2004; Van Etten 2009; Dunkerley 2012). When previously sparse areas of vegetation recover in response to precipitation, small animal populations subsequently increase, presumed to be spreading from population refuges in a spatial mosaic (Meserve et al. 1995; Ernest et al. 2000; Dickman et al. 2001; Berryman and Hawkins 2006; Haythornthwaite and Dickman 2006; Letnic and Dickman 2010; Thibault et al. 2010; Deguines et al. 2017; Pavey et al. 2017; Fischer et al. 2022). Refuges may be fixed geological or hydrological features (e.g. riparian corridors, rocky slopes, hill tops) or shifting portions of the landscape that reflect recent conditions (Milstead et al. 2007; Letnic and Dickman 2010; Pavey et al. 2017). However, we know little about the spatial scale and detectability of apparent refuges and, more broadly, the links between fine-scale spatial variation in the environment and animal communities (Pavey et al. 2017).

Spatial variation in pulses and persistence of plant growth occur at the landscape scale and at microhabitat scales (Lundholm 1976; Hamerlynck et al. 2002; Chesson et al. 2004; Li et al. 2009; Nano and Pavey 2013; Manning et al. 2020). Rainfall triggers a pulse of plant growth and release of nutrient reserves (the pulse–reserve model: Noy-Meir 1973; Reynolds et al. 2004). At the landscape scale, spatially variable rainfall interacts with soil formations (e.g. gullies and channels: Loik et al. 2004; Manning et al. 2020) that are created by aeolian or hydrological processes. Additionally, soils themselves vary in their water holding ability, further driving spatial variation in the resultant primary productivity (Chesson et al. 2004; Loik et al. 2004; Nano and Pavey 2013). Domestic grazing and fire in arid and semiarid landscapes also contribute to landscape-level variation in standing plant biomass (Letnic et al. 2005; Liu et al. 2011). At the microhabitat scale, there is local variation in soil composition, vegetative growth, seed banks and carbon stores that create mosaics of vegetation responding in a pulse of growth (Chesson et al. 2004; Nano and Pavey 2013). In addition, arid land plants often have species-specific responses to rainfall events (speed and nature of response), driving spatial variation at the microhabitat level (Lundholm 1976; Hamerlynck et al. 2002; Chesson et al. 2004; Li et al. 2009; Nano and Pavey 2013). Thus, the pulse–reserve process across a landscape with variation in abiotic and biotic features creates microhabitats of plant cover and composition as well as landscape variation in vegetative community.

Landscape-scale variation in vegetation in arid ecosystems gives rise to a spatial mosaic of animals (Ernest et al. 2000; Andrade and Monjeau 2014; Dalerum et al. 2017; Martínez et al. 2018; Marquart et al. 2020). Animal populations often rely on vegetation communities for success as vegetation provides resources or shelter (from the environment, predators or competitors: Haythornthwaite 2005; Letnic and Dickman 2010; Marquart et al. 2020). It is therefore unsurprising that many animals in arid landscapes increase in abundance following pulses in vegetation growth (Ernest et al. 2000; Stafford Smith and McAllister 2008; Thibault et al. 2010; Letnic et al. 2011; Kelly et al. 2013; Deguines et al. 2017; Burt et al. 2021; Fischer et al. 2022). Resulting spatial variation in vegetation, whether fixed or shifting, could create large-scale and small-scale refuges that allow animal populations to endure extreme and prolonged fluctuations in their environment such as drought conditions (Morton and Baynes 1985; Letnic and Dickman 2010; Letnic et al. 2011).

What remains unclear is whether ecological refuges are associated with vegetative resources. Does spatial variation in vegetation at the microhabitat level lead to detectable variation in the abundance of animals? On the one hand, many animals have very rapid life histories or high vagility, such that abundance can closely match local resource availability (Letnic and Dickman 2010), which would produce strong correlations between microhabitat patches and animal abundance. On the other hand, the responses of many animal populations occur only after a delay following changes in vegetation (Meserve et al. 1995; Ernest et al. 2000; Dickman et al. 2001; Thibault et al. 2010) or, alternatively, are ephemeral and difficult to detect without rapid repeated sampling (Maute et al. 2019; Fischer et al. 2022). In addition, some species may prefer open habitats, respond to microhabitat features other than vegetation (e.g. soil or rock crevices: Parmenter et al. 1989; Pierce et al. 1992; Berryman and Hawkins 2006; Letnic and Dickman 2010; Pulsford et al. 2017), or have adaptations that reduce their dependence on local variation in resources (e.g. retreating underground: Cloudsley-Thompson 1975; Lundholm 1976; Letnic and Dickman 2010). Moreover, trophic interactions may overwhelm the role of resource-driven patterns in determining animal abundance (Letnic and Dickman 2010; Deguines et al. 2017).

The aim of this study was to understand whether variation in habitat (rocky gibber plains versus alluvial clay plains) and microhabitat (local vegetation cover) creates spatial mosaics in animal diversity and abundance. Whereas many previous studies linking animal abundance to microhabitats have been specific in their taxonomic focus (Read 1987; Parmenter et al. 1989; Haythornthwaite 2005; Haythornthwaite and Dickman 2006; Thibault et al. 2010; but see Deguines et al. 2017) or have focused on temporal patterns (Read 1987; but see Ernest et al. 2000; Dickman et al. 2001; Haythornthwaite and Dickman 2006; Thibault et al. 2010), here we examine three major taxonomic groups of animals (invertebrates, small mammals and small reptiles) across multiple sites to examine spatial mosaics. We predicted that animal abundance and diversity would differ across the two habitats and increase with microhabitat vegetation cover.

Materials and methods

The study was conducted at the University of New South Wales Fowlers Gap Arid Zone Research Station in semiarid Australia. Fowlers Gap (31.09°S, 141.70°E, elevation ~181 m) covers 39,000 ha in semiarid NSW (Dunkerley 2013) and is characterized by chenopod shrubland and grassland plant communities. Fowlers Gap land has diverse soil formations and a complex usage history, including pastoral grazing since the 1880s. The property is split in half according to two dominant soil types (stony gibber plains and ranges, ‘Rocky’; flat alluvial clay plains, ‘Clay’), which creates large-scale habitat variation. It also offers fine-scale spatial variation in vegetation cover arising from standard arid ecosystem processes as well as the varied history of pastoral grazing. Thus, it is ideal for examining whether animal communities are linked to vegetation cover at small scales.

To address our aim, invertebrate and small vertebrate diversity and abundance were sampled in the two main geologic habitats of Fowlers Gap (Clay and Rocky) and across microhabitats that varied in vegetation cover and composition. Sampling occurred across six weeks in late austral summer (February–March 2022; one sampling season). Sixteen sampling sites were spread across the property, covering the two main habitats on Fowlers Gap equally (Fig. 1). At all sites, invertebrate sampling arrays were established. Six of the sites (three in each habitat) additionally had trapping arrays for small vertebrates. The six vertebrate sites were chosen based on availability of established traps. The 10 additional sites were selected to capture the range of apparent total vegetation cover on the property at the time of sampling, based on prior knowledge of the landscape. Thus, the sampling sites are not a random sample of total vegetation cover on Fowlers Gap or in each habitat. Three of the sites (vertebrate and invertebrate sampling arrays) were located inside grazing exclosures to include heavily vegetated microhabitats.

Fig. 1.

Map of UNSW’s Fowlers Gap Arid Zone Research Station showing trapping sites. Circles represent arrays (N = 16) of small pitfall traps that only captured invertebrates. Triangles represent arrays (N = 6) of large pitfall traps that captured vertebrates and invertebrates. Closed symbols are those in the alluvial plain clay soil; open symbols are those in the gibber rocky soils. Green boxes show three long-term grazing exclosures that were sampled.

Results

Animal captures

In total, during our summer trapping season, we captured invertebrates from 13 Orders and vertebrates from eight Families. Total number of invertebrates in small pitfalls was unrelated to habitat type when considering the number of Orders captured (F = 0.84, P = 0.38) or the total captures (F = 0.16, P = 0.69).

When looking at invertebrate captures in relation to vegetation cover, AIC model selection favored different vegetation cover variables depending on the response variable and type of pitfall array (Table 2). There were several positive relationships between invertebrate captures and vegetation cover; however, for most measures these relationships were not significant (Table 2, Figs 2 and 3). The one exception was that the number of spiders (Araneae) captured had a significant positive relationship with woody cover across the large pitfall arrays (Table 2, Fig. 3).

Table 2.AIC-favoured model for each response variable, with the best predictor and the ΔAIC for the next best model.

Pitfall	Response	Best predictor	ΔAIC	Test statistic
Small (N = 16)	Invertebrate Orders	Avg grass cover	0.339	z = −0.83, P = 0.41
	Total invertebrates A	Avg total cover	0.778	F = 1.90, P = 0.19
	Coleoptera A	Avg grass cover	0.574	F = 0.90, P = 0.36
	Aranae A	Avg woody cover	0.453	F = 0.50, P = 0.49
	Hymenoptera A	Avg total cover	0.722	F = 2.02, P = 0.18
Large (N = 6)	Invertebrate Orders	Avg woody cover	0.018	z = −0.22, P = 0.83
	Total invertebrates	Avg total cover	0.068	F = 0.61, P = 0.48
	Coleoptera	Avg woody cover	0.131	F = 0.27, P = 0.63
	Araneae	Avg woody cover	5.366	F = 10.74, P = 0.03
	Orthoptera	Avg total cover	1.649	F = 4.82, P = 0.09
	Vertebrate Families	Avg total cover	0.353	z = 1.19, P = 0.24
	Total vertebrate captures	Avg total cover	0.560	F = 4.30, P = 0.11
	Reptilia	Avg woody cover	7.230	F = 20.11, P = 0.01
	Mammalia	Avg grass cover	1.766	F = 2.84, P = 0.17

Test statistics are from generalised linear models (invertebrate orders, vertebrate Families) or linear models (all others). Values in bold indicate significance at α = 0.05.

^A Values are adjusted for site-specific trapping effort.

Fig. 2.

Best vegetation predictors for invertebrate captures in the small pitfall traps, including total captures (a and b) and the three most common Orders (c, d and e). Counts of invertebrates are an index adjusted for uneven trapping effort (number of captures/trap nights).

Fig. 3.

Best vegetation predictors for captures of invertebrates (a, c, e, g and i) and vertebrates (b, d, f and h) in the large pitfall traps. Included are the three most common Orders of invertebrates and two of the three vertebrate Classes captured.

Among vertebrate captures in large pitfalls, captures of reptiles had a significant, positive relationship with woody vegetation (Table 2, Fig. 3). The number of vertebrate Families, and total captures of vertebrates and mammals showed a non-significant positive relationship with vegetation cover (Table 2).

Animal community composition

When looking at the taxonomic groups captured across trap types, there were many more Hymenoptera (ant) captures recorded in the small pitfalls than the large pitfalls (Fig. 4a, c). In addition, there appeared to be more Araneae, Isopoda, Muridae and Scincidae captured in the Clay habitat and more Dasyuridae captured in the Rocky habitat (Fig. 4b, c, d). NMDS plots of community composition showed substantial overlap in site-based community composition from small pitfalls (Fig. 4e), but little apparent overlap from large pitfalls (Fig. 4g). However, for both datasets, community composition did not differ significantly between Clay and Rocky habitats, partly owing to the small sample size for large pitfall survey sites (ANOSIM, small pitfalls: R = −0.02, P = 0.47; ANOSIM, large pitfalls: R = 0.70, P = 0.10). Similarly, differences in community composition were not significantly related to differences in the combined vegetation measures across sites (Mantel, small pitfalls: R = −0.07, P = 0.70; Mantel, large pitfalls: 0.17, P = 0.27; Fig. 4f, h).

Fig. 4.

Community composition comparison between habitats and as a function of combined vegetation measures. Invertebrates from small pitfalls are shown (a) in entirety and (b) with the overwhelming number of Hymenoptera removed. Captures from the large pitfalls are shown separately for (c) invertebrates and (d) vertebrates. Comparison of community composition sampled with the small pitfalls (e) between the two habitats (NMDS plot: filled circles, clay; open circles, rocky) and (f) as a function of dissimilarity in combined vegetation cover. Comparison of community composition sampled with the large pitfalls (g) between the two habitats (NMDS plot: filled triangles, clay; open triangles, rocky) and (h) as a function of dissimilarity in combined vegetation cover.

Discussion

Animal abundance and diversity in arid ecosystems is often tied to temporal pulses in resources or large-scale variation in the landscape, thought to be driven in part by spatial refuges where animal populations persist during regional resource limitation. Here, we examined whether animal abundance and composition are linked with fine-scale spatial variation in habitat and microhabitat, such that these spaces might constitute refuges. We observed substantial variation in vegetation cover (i.e. microhabitat) and animal diversity across our sampled sites, yet animal diversity was not generally linked with our measures of habitat or microhabitat.

Our animal captures were suggestive of habitat-level differences in abundance of some taxa (e.g. spiders, beetles, dasyurid marsupials, and skinks), yet we found no significant difference in animal community composition between the two habitats. Previous study of dunnarts (Dasyuridae) at Fowlers Gap found that the species have little preference between habitat types in the area (Read 1987), which is in striking contrast to our notably higher number of dasyurids in the rocky habitat. When considering invertebrates across habitats, the large pitfalls captured many fewer invertebrates in the Rocky habitat compared with the Clay habitat, yet the relative abundance of the major Orders was similar (see also Oberprieler et al. 2019). The difference in overall captures is unlikely to be related to the low detectability of small invertebrates (particularly ants) in large pitfalls, as this trapping bias would impact captures in the two habitats equally. More likely, the increased abundance of marsupial predators in the large pitfall traps in the Rocky habitat led to the associated reduction in invertebrates. Indeed, captures from small pitfall traps had an overall similar number of captures between the two habitats but possibly fewer spiders and more beetles in the Rocky habitat compared with the Clay habitat. In arid lands more broadly, small animal communities have previously been found to vary across major habitats in arid lands (Ernest et al. 2000; Milstead et al. 2007; Dalerum et al. 2017; Martínez et al. 2018; Marquart et al. 2020). Thus, additional investigation is warranted before ruling out habitat-level differences between clay soils and gibber soils in our study area.

At a smaller spatial scale, we sampled across a patchy mosaic of vegetation microhabitats, which predicted the abundance of a subset of our taxa. Greater woody vegetation cover was associated with more spiders (Araneae) and reptiles, though these patterns should be confirmed with a larger sample. Woody vegetation in arid lands provides a stable resource of shelter and food. In particular, spiders in arid and woodland settings typically prefer woody habitat and litter (Whitehouse et al. 2002; Castro and Wise 2009). Similarly, many types of invertebrates (as well as small mammals) are more common near stable vegetation regardless of whether it is grass or woody vegetation (Read 1987; Haythornthwaite 2005; Marquart et al. 2020). Reptiles often benefit from woody debris and woody vegetation, as wood provides thermoregulatory opportunities, retreats from predators and increased termites for food (Morton and James 1988; Pike et al. 2011; Pulsford et al. 2017; Braun et al. 2024). Additionally, the soil around woody plants typically holds moisture for longer periods and acts as a source of water for invertebrates (Hamerlynck et al. 2002; Manning et al. 2020), providing a stable source of food for predators like spiders and reptiles. Thus, our results suggest that woody chenopod shrubs form the foundation of microhabitat mosaics for certain taxa and potentially support population persistence as relatively fixed refuges during droughts.

Despite the suggestive patterns seen in spiders and reptiles, other measures of animal abundance and communities were unrelated to microhabitat vegetation cover. This may reflect true limitations on the extent to which vegetation cover provides population refuges. For example, preferences and sensitivities with respect to habitat and microhabitat are often species-specific and may not be linked to vegetation cover per se (Haythornthwaite and Dickman 2006; Pavey et al. 2017; Lu et al. 2021; Hernandez-Santin and Fisher 2022). In arid Australia, dasyurids show species-specific responses to rainfall and vegetation, with captures sometimes negatively related to microhabitat grass cover (e.g. preferring recently burned sites) or seemingly unrelated to habitat measures from multiple spatial scales (Masters 1993; Haythornthwaite and Dickman 2006; Hernandez-Santin and Fisher 2022). Similarly, previous work at Fowlers Gap performed across a smaller spatial extent than the present study suggested that captures of small dasyurid marsupials are linked to microhabitat features such as vegetation height and density of soil cracks but not to general habitat characteristics (Read 1987). Among invertebrates, percentage grass cover has been linked strongly to populations of herbivores and omnivores but not predators (such as spiders) and detritivores (Marquart et al. 2020; Fischer et al. 2022; Heimbuch et al. 2023). In addition, invertebrate populations in semiarid grasslands have been found to be highly sensitive to plant quality (i.e. leaf nutrient content) and new growth after fire, with less of a role for total biomass (Lu et al. 2021; Heimbuch et al. 2023). Moreover, diversity and abundance of some taxa may also be influenced strongly by predation, competitive exclusion, and their interaction with vegetation (Brown and Munger 1985; Letnic and Dickman 2010; Atwood et al. 2011; Dickman 2014; Terborgh 2015; Deguines et al. 2017).

On the other hand, true links between vegetation cover and animals are challenging to detect in arid lands due to the unpredictable and ephemeral nature of resource availability. In our study, Fowlers Gap received rainfall in the weeks before sampling, and there was already a high abundance of grass and forbs present when vegetation cover was estimated. Although animals in arid lands typically respond to ‘greening’ of vegetation (Hobbs et al. 2008; Letnic and Dickman 2010; Fischer et al. 2022), the timing of the response is variable and we sampled over a short timescale. Vertebrate population numbers exhibit a substantial delay to greening (~1 year) (e.g. Meserve et al. 1995; Dickman et al. 2001; Thibault et al. 2010). Thus, current vegetation cover, particularly fast-growing grasses and forbs, is unlikely to be a good predictor of concurrent vertebrate communities (Dickman et al. 2001). On the other hand, invertebrate populations respond quickly to rainfall (<1 month) and grass growth, yet the effect may be detectable only in specific, limited time windows after rain (Fischer et al. 2022). For example, at Fowlers Gap, invertebrate captures have been found to increase dramatically within two weeks after rainfall, but numbers of only some taxa remained elevated three months later (Maute et al. 2019). These challenges in detecting links between microhabitat characteristics and animal abundance may explain why only woody cover had explanatory power in our study. Persistent vegetation such as woody chenopod shrubs responds to rainfall more slowly (Nano and Pavey 2013; Manning et al. 2020) such that populations of animal species that are sensitive to woody cover may be more likely to maintain numerical associations with woody cover. Importantly, invertebrate taxa that use woody vegetation as an important microhabitat are less likely to be captured with pitfalls and instead require bush beating (e.g. Martinez et al. 2021). Recording invertebrate and vertebrate assemblages at a finer taxonomic scale (e.g. genera or species) and across multiple sampling periods could help tease apart the temporal dynamics and niche variation underlying potential links between population abundance and vegetation.

Refuges are also challenging in a spatial context. Populations of small animals may require a larger spatial extent of favourable habitat than the scale of microhabitats that we sampled. In addition, by definition, we may expect refuges to be rare (Milstead et al. 2007), so they may be difficult to detect in the standard ecological sampling framework. Finally, it is possible that refuges are not places where taxa persist more abundantly during periods of low resource availability but rather places where growth can happen rapidly under the right conditions. Identifying animal refuges in arid settings may often require far greater spatial and temporal scales of sampling than are typically possible in most studies (Ernest et al. 2000; Haythornthwaite and Dickman 2006).

Overall, our results suggest that refuges were not strongly apparent in our study. However, the implications of our findings for arid-zone mosaics and refuges should be viewed in light of the limited temporal scale of the study. Spatial mosaics in animals existed in our study but linking them to driving forces may require understanding vegetation and animal communities over time. Multiple well-timed trapping events are required to gain further knowledge of the relationship between vegetation cover (habitat and microhabitat scales) and animal community composition, abundance, and diversity. This type of monitoring will provide insight in the future as Fowlers Gap has recently removed grazing sheep and will be experiencing altered grazing pressures. Diversity in arid landscapes is often strongly associated with grazing pressure in arid lands (Debano 2006; Liu et al. 2011), so changes to plant and animal communities will be expected. Hopefully, the current study will provide a baseline for future surveys of diversity and abundance following the removal of domestic grazers.

Supplementary material

Supplementary material is available online.