Ecological Drivers and Ant Community Structure in Neotropical Biomes

We sampled ants along ten transects in each biome that were spatially separated by at least 1 km across the study (n = 60). We measured both species richness and assessed two niche dimensions - their diet and foraging habitat strata - related to five different liquid food resources at three habitat strata. Transects were 740 m long and presented 75 sampling points, separated by 10 m from each other. We placed 50 ml fisher tubes containing one of five liquids resources soaked in 5 cm cotton balls. Liquid resources were made by a solution of (distilled water/volume): 1% sodium (NaCl), 20% sugar (made with sucrose), 20% amino acids (made with unflavoured whey protein isolate) and 100% lipids (extra virgin olive oil) and 100% distilled water, as a control. Along the transects, we alternated one of the five liquid baits at one of three habitat strata (arboreal, epigaeic and subterranean, see details in Lasmar et al., 2023) totalling 4,500 baited tubes. The baited tubes placement started at 7:00 am for each transect, except in the Caatinga biome due to morning rainfall events. All baits remained operating for 3 h. Since distilled water baits were infrequently visited by ants (see Lasmar et al. 2023), we excluded such baits when categorizing the niche of ants. We, therefore, had a combination of 12 niche aspects related to the diet and foraging habits of ants across habitat strata (4 bait types × 3 habitat strata) and five replications of each of niche aspects per transect.We identified all collected ant workers into genera following Baccaro et al. (2015), and whenever possible to species and to morphospecies using the relevant literature (data in Lasmar et al., 2022) and/or matching individuals with the ant reference collection at the ‘Laboratório de Ecologia de Formigas’ of the Universidade Federal de Lavras (UFLA) and Entomological Collection Padre Jesus Santiago Moure of Universidade Federal do Paraná (UFPR). Voucher specimens are deposited in the reference collection of both collections at UFLA and UFPR.<br><i>Measuring ecological drivers</i><i>Net primary productivity and Habitat Heterogeneity</i>Net primary productivity (NPP) was obtained from MODIS NPP data, which represents the annual average of 2000 – 2015 years (MOD17, from NASA Earth Observation System). This is available in the repository at the University of Montana (www.ntsg.umt.edu/) at 1 km spatial resolution. The data obtained was an improved version of MOD17, which cleans cloud-contaminated pixels and considers the difference between gross primary productivity and autotrophic respiration (Zhao and Running 2010).In a parallel transect, 20 m apart from each baiting transect, we measured habitat attributes at five sampling points 50 m apart from each other to create a heterogeneity index. Details are in Appendix 1. As productivity is generally correlated with habitat heterogeneity, we checked for collinearity using the "<i>corrplot</i>" function from the 'corrplot' package in R (Wei and Simko, 2021). We found that more productive regions were also more heterogeneous (R = 0.65). Therefore, we solely considered net primary productivity data for the analyses as it better encompasses the area of our transects.<br><i>Contemporary and past climate and stability over geological time-scales</i>To accurately represent the current and past climate conditions, we used data from various sources. For the current climate, we extracted for each of our 60 transects the mean annual temperature (°C), annual precipitation (mm), temperature seasonality (standard deviation in °C of annual mean temperature) and precipitation seasonality (coefficient of variation of annual precipitation) from Worldclim2 dataset with a spatial resolution of 1 km grid cells and measures between 1970 to 2000 (Fick and Hijmans 2017). We ensured that our current climate data from Worldclim2 closely matched the weather data from the nearest climatic station measured from 1970 until the date we carried out the samplings of each biome (see Lasmar et. al., 2021a). This alignment indicated that our sampling was not conducted during extreme weather conditions and that the Worldclim2 data accurately represented both the local weather during at the time of sampling and the long-term climate trends.For past climate data, we obtained the current corresponding ecological drivers (mean annual temperature, annual precipitation, temperature seasonality and precipitation seasonality) at various time points. These time points included the last glacial maximum (21 ka; Karger et al. 2017) at 1 km of resolution, the last interglacial in the Pleistocene (<i>c</i>. 130 ka; Otto-Bliesner et al., 2006), the early Pleistocene (<i>c</i>. 787 ka. Brown et al., 2018), and the Pliocene (<i>c</i>. 3.3 Ma; Dolan et al., 2015) - all last three at 5 km of resolution. All past climate data were downloaded on PaleoClim.org (Brown et al., 2018). To assess the collinearity between current and past climate variables, we employed the <i>corrplot</i> function from the ‘corrplot’ package in R (Wei and Simko, 2021). All current climate variable representations were highly correlated with their past correspondents (i.e., r &gt; 0.80; Appendix 1, Fig. S2-S5), indicating that at least in a range of <i>c</i>. 3.3 Ma, geographical variation in temperature, precipitation and their seasonality of our sampled regions remained almost the same. Thus, we solely considered current climate data for the analyses.For each transect, we also obtained the climatic stability over geological time-scales data from Herrando-Moraira et al. (2022) study, with 5 km of resolution and corrected for the collinearity of climatic variables and the sea level variation across history. The climatic stability index was created by extracting the standard deviation (SD) as an estimate of the amount of variation of 14 climatic variables available in PaleoClim.org across 12 time periods, from Pliocene (3.3 Ma) to the present. Then, it was summed the SD values of all 14 variables and this value was normalized from 0 to 1 (i.e., 0 completely stable; 1 most unstable).