Disturbances are natural or anthropogenic processes removing biomass from communities (Grime, 1977). They modify local abiotic properties, habitat structure and resource availability creating important environmental filters for biota, thus shaping community assembly, functional traits, ecological interactions, and functions (Bond and Keeley, 2005; Díaz et al., 2007). Open ecosystems worldwide such as savannas and grasslands have a close evolutionary relationship with grazing and fire, and therefore present not only disturbance-adapted but also disturbance-dependent communities (Veldman et al., 2015). In these ecosystems, plant conservation can be suitably coupled with traditional management (e.g., grazing, excepting intense forms, prescribed burnings), while disturbance suppression policies usually lead to typical plant diversity loss by competitive interactions (Abreu et al., 2017; Ferreira et al., 2020; Koch et al., 2016). Prescribed burning is a commonly applied management tool to improve forage value, to prevent wildfires, and ultimately to increase landscape heterogeneity and benefit biodiversity where grazing is low or absent (Bond and Keeley, 2005; Fernandes et al., 2013; Morgan et al., 2020) Nevertheless, to foster efficient conservation strategies in open ecosystems, it is fundamental to understand the effects of disturbances and their suppression on a wide-range of organisms, since, among animals for example, responses can be taxa dependent (Pastro et al., 2011).

Here we performed a controlled patch burning experiment to shed light on how prescribed burnings in long-term disturbance-suppressed grasslands affect ant communities and their functions in South Brazil. Ants are a globally dominant faunal group with sessile colonies, performing outstanding ecological roles (Folgarait, 1998). Ants display a large range of feeding strategies, removing widely dispersed aboveground resources such as plant diaspores, nectar, dead animal remains, and leaves. By transporting these resources to their nests, ants help redistribute nutrients in the ecosystems (Farji-Brener and Werenkraut, 2017). In grasslands, ants nest essentially belowground, where direct colony mortality by fires is negligible (Arnan et al., 2006; Debano, 2000). Nevertheless, ant communities are commonly affected by fire-induced habitat changes, such as in structure, microclimate, and resource availability (Andersen, 2019). For example, post-burnt grassland sites usually present higher habitat openness (e.g., suppressed vegetation, high sun exposure, soil temperatures, and low moisture retention) and simpler substrate (e.g., suppressed litter, high bare soil proportion) for ant locomotion than unburned sites (Podgaiski et al., 2014). Such conditions could eventually trigger species and trait-based community assembly through an environmental filtering process (Bishop et al., 2021; Wiescher et al., 2012), besides changing the community foraging patterns (Parr et al., 2007). Also, as plant diversity has been widely proposed as a proxy of food diversity for ants (Dröse et al., 2021), it is expected that the pace of post-fire vegetation regeneration would drive the diversity patterns of foraging ant communities (Bishop et al., 2021).

We propose hypotheses and predictions on how fire would indirectly affect ant species composition, community body size, species richness, and seed removal by ants.

• (1)

  Fire-induced habitat openness trigger ant species composition changes. We predict a turnover of ant species post-fire. Species more prone to forage under open habitat will be favored, while those with preferences for more shady environments will be disfavored in burned grassland (Andersen, 2019; Farji-Brener et al., 2002; Lessard and Andersen, 2019).

• (2)

  Due to the more simplified and sun-exposed environment, burned grasslands benefit ant species with larger size. Among functional traits, ant body size is related to mobility and desiccation resistance (Farji-Brener et al., 2004; Kaspari and Weiser, 1999). Large-bodied foragers present faster mobility in less complex habitats relative to more complex ones (Farji-Brener, 2004b; Parr et al., 2003), and also show more heat tolerance (e.g., UV radiation incidence) than small ants (Kaspari et al., 2015; Kaspari and Weiser, 1999). Therefore, we predict a higher ant community mean body size in burned than unburned grasslands.

• (3)

  The high plant richness of burned grasslands favors the coexistence of a great number of ant species. Fires usually increase grassland plant diversity by reducing competition for light and space among plants, favoring growth, reproduction and early flowering of plant species with less competitive capacities (Fidelis et al., 2012; Fidelis and Blanco, 2014; Joner et al., 2021; Overbeck et al., 2005). By considering plant richness as a proxy of resource diversity for ants and that resource diversity allow coexistence of foraging ant species (Dröse et al., 2021), we predict a positive association of plant and ant richness, and a higher ant richness in burned than unburned grasslands.

• (4)

  The simplification of the habitat post-fire makes it easier for ants to forage and find seeds. The suppression of vegetation cover by fire opens the habitat, facilitating ant locomotion and activity, usually resulting in greater resource finding (Dolabela et al., 2020; Gibb and Parr, 2010; Radnan et al., 2018). Thus, we predict higher rates of seed removal by ants in burned than unburned grasslands, and that such rates would be predicted by ant activity.

We sampled a total of 4316 ants belonging to 29 genera, and 57 species. Camponotus crassus Mayr, 1862, Brachymyrmex sp., Camponotus fastigatus Roger, 1863, Holcoponera striatula Mayr, 1884, and Ectatomma brunneum Smith, 1858 were the most frequent species in the study (Supplementary Material 2). The total number of sampled species varied with season of sampling, with lower richness accumulated in the wintertime (6 months post-fire; Fig. 1). Plots from different treatments shared a large proportion of species (i.e., about 60% of the total community of each time), but burned plots accumulated higher total species richness than controls (i.e., 11.4%, 35% and 27.6% more species at 1-, 6- and 12-months post-fire respectively; Fig. 1). Burned plots also accumulated more unique species than controls.

Fig. 1.

Venn diagrams showing the total number of ant species (s), the number of species on each treatment (bold numbers), and the number of exclusive and shared species (italic numbers) between treatments for each sampling occasion (Pre-fire, 1-, 6- and 12-months post-fire). Burned plots in yellow, controls in grey.

(0.17MB).

Ant species composition did not significantly vary between treatments: either pre-fire (F1,18 = 0.63; p = 0.84), 1 month (F1,18 = 0.92; p = 0.55), 6 months (F1,18 = 0.5; p = 0.92), and 12 months post-fire (F1,18 = 0.6; p = 0.86). Ant community body size also did not vary (Table 1A). Confidence intervals on the effect size value greatly overlapped with the null effect value (Fig. 2A), indicating no fire effects on CWM body size at any sampling occasion.

Fig. 2.

Mean fire effect size (Hedges' g) and bootstrapped 95% confidence intervals on (A) CWM ant body size, (B) ant species richness, and (C) seed removal rates in each sampling occasion (pre-fire, 1-, 6-, and 12-months post-fire). Treatment effect is significantly positive or negative when intervals do not overlap zero.

(0.25MB).

At the local scale, mean species richness varied between burned and control plots (Table 1B). We detected significantly positive fire effects on ant richness after 1 month (Hedges' g = 1.20) when burned plots accumulated 10% more species. At this time, each burned plot presented on average 2.57 (18.7%) more species than controls. We detected stronger fire effects after 12 months (Hedges' g = 2.28) with 22% more species at total. At this time, burned plots had on average 3.3 (25.9%) more species per plot. No significant fire effects were detected for 6 months post-fire (Fig. 2B). The SEM testing the hypothesis that ant richness increased after fire (12 months) due to direct effects of the diversity of resources was validated (Fisher's C = 4.085; df = 2; p = 0.13), and all single relationships were significant (p < 0.05; Fig. 3A).

Fig. 3.

Structural equation models (SEM) testing indirect fire effects on (A) ant species richness through changes in plant species richness at 12 months post-fire, and (B) seed removal rates (%) through changes on ant abundance of selected species (see Supplementary Material 2) at 1-month post-fire. The percentage number within the boxes indicate the conditional R-squared of the relationship. Yellow arrows represent positive relationships (**p < 0.01; *p < 0.05), and white arrows non-significant relationships (p > 0.05). In the scatterplots, black circles represent burned plots and white circles control plots.

(0.17MB).

Mean seed removal rates varied between burned and control plots (Table 1C). We found positive fire effects on removal rates after 1 month (Hedge' g = 1.09), when burned plots presented 19,76% more seeds removed per plot than controls. No fire effects were detected for other sampling occasions (Fig. 2C). The SEM testing the hypothesis that seed removal at this time was influenced by ant activity was validated (Fisher’s C = 3.80; df = 2; p = 0.15), but none of the single linear relationships were significant (p > 0.05; Fig. 3B).

We experimentally burned patches in a long-undisturbed grassland to quantify fire effects in ant communities and their foraging behavior. Our results suggest that patch burnings do not harm grassland-adapted ant communities. On the contrary, by reducing aboveground biomass and reassembling plant communities, fire creates temporary resource-rich habitats for these communities that subtly increase species richness and resource use.

We had initially hypothesized that patch burnings would change local ant composition and community body size by altering grassland openness. Nevertheless, our experimental results did not show changes in overall ant species occurrences across burned and control sites. This probably happened because ant colony mortality directly caused by fire was negligible (as in Arnan et al., 2006; Frizzo et al., 2012), and the resprouting and recover of the grassland vegetation after fire was extremely fast (Joner et al., 2021). Community body size also did not shift in the direction we had predicted, as would be expected if habitat simplification and higher sun exposure had locally selected groups of ant species with larger body sizes. While larger ants may indeed have advantages against desiccation in a harsh post-fire environment (Bujan et al., 2016), they can also be more detectable to visual predators, such as birds, that are usually very abundant post-fire (Beal-Neves et al., 2020). This may be potentially masking the trait-based patterns, which should be clarified in further studies. Moreover, Farji-Brener, 2004b also emphasize that inconsistent associations between the rugosity condition of an environment and ant body sizes may happen in manipulative experiments because such associations operate better at evolutionary rather than ecological scales. Nevertheless, our results concur with other studies in open ecosystems worldwide (Calcaterra et al., 2014; Lopes and Vasconcelos, 2008; Hosoishi et al., 2015; Izhaki et al., 2003; Parr et al., 2003) indicating a highly resilient, or resistant, ant fauna associated to fire disturbances in these types of ecosystems, even after long-term absence of disturbances (Parr and Andersen, 2008).

As predicted in our hypothesis, burned patches supported higher local ant richness (at the patch-scale), and also accumulated more species (at the study site-scale) than unburned sites. While a recent meta-analysis did not show global significant fire effects on ant diversity in open ecosystems worldwide (Vasconcelos et al., 2017), several studies like ours have indeed found ant diversity to increase after fires, such as in the Brazilian, African, Australian and Mediterranean savannas/grasslands (Andersen et al., 2014; Bishop et al., 2021; Maravalhas and Vasconcelos, 2014; Parr et al., 2004). Shortly after fire (1 month) this pattern was weaker and probably influenced by an increased detectability of resident species foraging in a simplified environment (Melbourne, 1999) since despite a significant increase, we found only 10% more species. Nevertheless, one year after fire the effect was more robust, when burned patches reached about 25% more species at both scales (patch and site). Ant species aggregations on burned patches at this later time could be a result of increased nesting processes of different species along the former year. Moreover, this may also indicate that species that nest in the unburned matrix are foraging within the burned patches. Although most ground ant species forage just in the immediate vicinity of their nests (some centimeters to a few meters; (Peeters and Ito, 2001)), some medium and large-bodied species can indeed present high foraging ranges, and directly select for high-quality food spots, as the burned patches (Ronque et al., 2018). Ant richness clearly mirrored the increment in plant richness in burned patches, a possible source of resource diversity. The species-rich regenerating vegetation can be very attractive to ants by offering a suit of exploitable resources both in a direct (floral and extrafloral nectar, fruits, seeds, palatable leaves) and indirect way (honeydew from trophobionts and insect prey living on the vegetation), benefiting generalist species, and also those with specific feeding habits. High resource availability could also relax interspecific competition among ants fostering resource partitioning and coexistence of more local ant species (Ribas et al., 2003; Senior et al., 2013).

Seed removal by ants was about 20% higher in freshly burned patches in comparison to unburned sites. Ant foraging efficiency has been frequently linked to habitat simplification at the ground level, where a simpler and warmer environment could enable higher mobility to ants, which would find and remove resources more quickly (Beaumont et al., 2011; Gibb and Parr, 2010; Paolucci et al., 2016). For example, da Silva et al. (2020) found ants interacting with more individuals of a plant species bearing extrafloral nectaries (Chamaecrista repens) in recently burned grasslands in comparison to grasslands longer without disturbances, and ant-visiting densities were positively predicted by bare soil proportions. The removal of seeds, although increased after fire, was not linked to the abundance of potential seed harvesters, as expected (Fig. 3). Similarly, Parr et al. (2007) detected higher rates of seed removal post-fire in an Australian tropical savanna, which was primarily driven by some key ant species (Iridomyrmex) with highly patchy distributions. Probably, our removal rates were also influenced by particular key species (e.g., Pogonomyrmex, Pheidole). However, our experimental approach did not allow us to assess this directly. Nevertheless, here we do show that fire-induced habitat changes facilitate seed removal by ants, indicating an efficient use of the available resources, especially under highest fire-induced habitat simplification.

We showed that prescribed patch burnings in fire-prone grasslands promoted ant richness, its foraging activity, and potentially fostered the maintenance of ant-mediated ecosystem processes. Our experimental approach, considering small-scale patches distributed within an unburned matrix resembles a mosaic landscape as it occurs in some regions in South Brazil, where fire spreads rapidly and heterogeneously through a wavy relief with dry/humid areas that vary regarding available flammable biomass. Our study, combined with others, may serve as a basis for conservation management decisions, endorsing the importance of disturbances in grasslands to keep and increase plant and ant patch diversity. Future studies should complement our results by investigating fire-induced environmental filtering on ant community assembly and their associated roles under varying burned patch size, fire intensity and frequency in subtropical grasslands, to help more fully subsidize management actions.

LRP and MSMJ contributed to the study conception and design. Material preparation and data collection were performed by LRP and CPRF. Data organization and analysis were performed by GGB and LRP. The first draft of the manuscript was written by GGB and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

This work was supported by CAPES, Brazil (Finance code 001) that provided Master scholarships to GGB, and Post-doc grants to LRP. CNPq (Brazil) provided Productivity Grant to MSMJ (grant number 311298/2019-2).

All data will be made available by the authors by request.