Aim: Ant communities are believed to be structured by competition, with dominant species competitively excluding subordinates (the dominance-impoverishment rule). However, a high number of seemingly similar species coexist, possibly due to interspecific trade-offs. Here, we examine the evidence for the dominance-impoverishment rule across a broad latitudinal gradient and explore whether trade-offs explain coexistence within and among ant communities.

Location: 40 sites in 19 countries across Europe, western North America and northern South America.

Taxon: Formicidae.

Methods: We conducted 2-h baiting experiments at each site. Three dominance scores were calculated for each species at each site where it occurred. We then examined the relationship between ant dominance and diversity and tested for the generality of three trade-offs (dominance-discovery, dominance-thermal tolerance and dominance-generalism) within and among ant communities along with the possible effects of environmental variables on these trade-offs.

Results: We found no support for the dominance-impoverishment rule. Instead, overall species richness at baits was positively correlated with the number of dominant species and exhibited a unimodal relationship with the relative abundance of dominant ants. Moreover, we found little consistent evidence for the three interspecific trade-offs.

Main conclusion: Although total species richness at baits is positively correlated with species richness of dominant species and, to a point, increasing worker numbers of dominants, trade-offs among species do not appear to shape broad-scale patterns of coexistence among ants. Species richness declines only when the numbers of dominant workers are very high. Together, these results suggest that while trade-offs and the dominance impoverishment rule might promote coexistence or shape ant communities in some locations, the evidence for their being general across communities is scant.

We conducted baiting experiments at 40 sites of 20 x 20 m (each consisting of four 5 x 5 m subplots) during the daytime between June 29^th and November 11^th 2016 at 40 sites: 20 in Spain, seven in France, five in Germany, three in Denmark, four in the United States, and one in French Guiana (Fig. 1). Sites were in areas with known ant activity or where ant fieldwork had previously been carried out and spanned a range of ecological biomes, including temperate forests, deserts and xeric shrublands and rainforests. Sites were separated by at least 100 m. We designed the experiments similarly to classic experiments in ant ecology (Fellers, 1987; Savolainen & Vepsalainen, 1988; Savolainen, Vepsäläinen & Wuorenrinne, 1989; Andersen, 1992; Perfecto, 1994; Cerdá et al. 1997; Sanders & Gordon, 2003). Specifically, we chose five different resources (canned tuna in water, untoasted sesame seeds, 20 % sugar water solution, 1 % saltwater solution and tap water) to attract diverse species at each site. Approximately one teaspoon of solid resources and 2.5 cm diameter cotton balls soaked in the liquid resources were placed on individual ~6 cm diameter plastic discs in the middle of each subplot in a pentagonal shape, equidistant from the plot boundaries and approximately 20 cm from each other. Twenty baits were deployed at each site: One bait per bait type (5) per subplot (4). After deploying the baits, the numbers and identities of ants present were recorded after 5, 15, 30, 60, 90 and 120 minutes. Ground temperatures were measured during each observation using a handheld infrared thermometer (Raytek Raynger ST). Ants were identified to species or morpho-species either in the field or in the lab. 

We extracted four environmental variables from online databases for each site. Mean annual temperature (MAT) and annual precipitation (AP) data were extracted from the 1970-2000 average WorldClim2 dataset at a resolution of 30 arc seconds (Fick & Hijmans, 2017). We extracted monthly normalised difference vegetation index (NDVI) values for each site during the month when sampling occurred from Moderate-Resolution Imaging Spectroradiometer MOD17, 30-arcsec data (Didan, 2015). We estimated actual evapotranspiration (AET) using Turc’s formula (Turc, 1954; Kluge et al., 2006; Sanders, Dunn, Fitzpatrick, Carlton, Pogue, Parker & Simons, 2009), where AET = P/[0.9 + ( P/L)²]^1/2 with L = 300 + 25T + 0.05T³, P = annual precipitation and T = annual mean temperature.