Neither ant dominance nor abundance explain ant-plant network structure in Mexican temperate forests

Ant-plant network sampling

MNP’s network data was taken from a previous study by our research group (Lara et al., 2020). This study reported the ant-plant networks obtained by monthly sampling (from June 2015 to July 2016) conducted in conserved and perturbed oak forest sites. The network data derived from the two vegetation types present in the study area (oak forest and grassland) was used for the analyses described below. FBSP’s networks were acquired in 2016 by monthly sampling (from January to December 2016) of each site’s ant-plant interactions. Both of these studies were carried out during the dry and rainy seasons and were conducted with the following protocol. To avoid bias, observers were trained by a single team member (M. Cuautle) before initiating the data collection. Walking censuses were undertaken from 08:00 h to 13:00 h on days with mild weather conditions, commencing from a different paired transect and vegetation type to avoid order effects and the confounding effect of detection of interactions in the field. The same observer conducted the day’s census, taking around 1 h in each transect. Observers recorded any type of interaction between individual ants and plants (i.e., foraging for floral, extrafloral nectar or honeydew), at a distance of no more than 10 m from the centerline of the transect. In trees, ant-plant interactions were recorded up to 1.5 m above ground level. The ants were collected with an entomological aspirator and stored in Eppendorf tubes with 70% alcohol. In addition, samples were collected from the plants with which the ants interacted. Both ant and plant specimens were identified using taxonomic keys. R. Acosta-Pérez from the Universidad Autónoma de Tlaxcala, Botanical Laboratory identified the plants, and G. R. Pérez-Toledo made most of the ant determinations up to species level (using the taxonomic keys of Baroni, 1978; Mackay & Mackay, 1989; Francoeur, 1973; MacKay, 2000; Longino, 2009; Cuezzo & Guerrero, 2011; Stockan et al., 2016). W. Mackay from the University of Texas, Laboratory for Environmental Biology Centennial Museum, determined the ant Camponotus rubrithorax.

Ant-plant network metrics

For each ant-plant community, interactions were summarized as a bipartite matrix, consisting of cells filled with the pairwise interaction frequency of all ant and plant species. Separate analyses were performed for each vegetation community (oak forest and grassland) in both areas (MNP and FBSP). Given the similarities found in the oak forest and the grassland of the MNP (three of six shared ant species and eleven of thirty shared plant species), as a second approximation, data derived from both communities were pooled. This approximation increased the sample (number of ant species) used for the correlation analysis, which is worthwhile because a larger sample size can augment the test’s statistical power (Zar, 1996). The results of the analysis performed on the pooled data from the MNP communities are presented in a Table S1. The following complementary network metrics were calculated; (1) nestedness, (2) web asymmetry, (3) specialization asymmetry, (4) species strength, and (5) identification of the core and peripheral species in each ant-plant network. (1) Nestedness quantifies the degree to which interactions of specialized species are subsets of interactions of the more generalist species in the networks and was calculated with the NODF estimator of the ANINHADO software (Guimarães Jr & Guimarães, 2006; Almeida-Neto et al., 2008). The topology of a network was considered nested when the observed NODF value (NODF) was higher than predicted by the null model Er (NODF(Er)), with 1,000 randomizations for each network (Guimarães Jr et al., 2006). In the null model, (Er) interactions are randomly assigned (Bascompte et al., 2003). Web asymmetry, species strength, and specialization asymmetry were estimated with the bipartite library (Dormann, Fruend & Gruber, 2018) in the statistical software package R (R Development Core Team, 2014). (2) Web asymmetry is the difference between the number of ant and plant species in a network divided by the total number of species in that network, where positive values indicate higher-trophic-level species, while negative values denote lower-trophic-level species (Blüthgen et al., 2007). (3) Specialization asymmetry, based on the degree of specialization of each species (d′), does not vary significantly across networks of different dimensions. Since the mean d-value for the lower trophic level (plants) is subtracted from that of the higher level (ants), positive values indicate a higher specialization of the higher trophic level (Dormann, Fruend & Gruber, 2018). The degree of specialization can be higher or lower than expected, given the web asymmetry (Blüthgen et al., 2007). To determine if this was the case, we estimated the significance of web specialization asymmetry with a Monte Carlo procedure in which 100 random matrices were generated using the null model (Patefield algorithm), where marginal totals were identical to those of the observed network (Anjos, Dáttilo & Del-Claro, 2018). To define the role of each interacting species within the networks, we used (4) species strength (ST) that is the importance of a species for the entire network, i.e., the sum of the dependencies of all interaction partners of that species (Bascompte, Jordano & Olesen, 2006). Lastly, to identify (5) the core and peripheral species in each ant-plant network, we used the equation: ${\displaystyle G_i=\left(\frac{k_i-\overline{k}}{\sigma k}\right)}$ where G_i is species centrality; k_i = average number of links involving an ant species i; $\overline{k}$ = average number of links involving all ant species in the network, and σ_k is the standard deviation of the number of links of all ant species. G_c values >1 were core species, and species <1 were peripheral species. Core species interact with virtually all species in the network, while peripheral species have fewer interactions with other species of the same trophic level (Dáttilo, Guimarães Jr & Izzo, 2013).

Nestedness measure correlations

Nestedness measures that represent the ant’s position in the network were used to assess whether dominance hierarchy determines the nestedness scores in the ant-plant network. (1) Nestedness ranking (Nra) (Dáttilo et al., 2014b), that is, the species of ants in the network were organized according to their degree of nestedness (i.e., the number of links per species), and ordered from highest to lowest. (2) Nestedness contribution (Nc). To obtain this parameter, we used the equation Nc = (100 − IT)/100, where IT is the idiosyncratic temperature, species with high IT have an atypical pattern of interaction compared with other species of their same trophic level (Vilà et al., 2009). Because generalist species interact with many other species, their interaction pattern is similar to other species at the same trophic level, resulting in lows ITs and highs Ncs; the contrary happens with the specialist species. ITs were calculated with ANINHADO software (Guimarães Jr & Guimarães, 2006; Almeida-Neto et al., 2008). (3) Nested rank (Nr) orders species according to their generality, which is measured as their position in the nestedness matrix. A generalist will interact with more species and thus have a rank closer to 0, while specialists (and rare species) will have higher ranks (closer to 1) (Alarcón, Waser & Ollerton, 2008). Nested ranks were calculated using the bipartite library (Dormann, 2011; Dormann, Gruber & Fruend, 2008; Dormann, Fruend & Gruber, 2018) in the statistical software package R (R Development Core Team, 2014). (4) Expected frequency was calculated only for FBSP. The sample numbers from the MNP were too low to make this approximation worthwhile, as the ant’s expected frequencies would not differ from their degree of nestedness. A plot was constructed of the frequencies of the grades (i.e., the number of links) for both types of vegetation in FBSP’s network. Ideally, the frequencies plot will have a smooth, continuous estimate of frequencies at each grade. To obtain that continuous estimate, we used a nonparametric approximation: kernel density estimation. Based on a set of observations, this technique estimates the density function from which the observations may have been derived. In this case, most of the sets of frequencies taken from the observations of FBSP’s oak forest and grassland are close to zero making it difficult to obtain an accurate density estimate with this technique. To overcome this issue, we used the transformation kernel density estimator. It spreads the observations out, finds their density estimate, and then reverts them to the original scale, resulting in a distribution that fits the observed density. From here, obtaining a smooth pattern for the frequencies is straightforward, since both the frequency and density functions are the same except for a constant of proportionality. A log transformation was used to implement this methodology. This procedure is well explained and justified in Wand & Jones (1995). Subsequently, these four nestedness measures were correlated with the abundance, DI, and MDI using a Spearman correlation.

All data collection and analysis of the FBSP sampling was conducted for the present study. Whereas for the MNP, the ant-plant matrix data was compiled from Lara et al. (2020), and the dominance hierarchy ranking was taken from Castillo-Guevara et al. (2019). However, the ant abundance calculation and data analysis of MNP were carried out for this study.

Ant-plant network MNP

In MNP’s oak forest, 18 interactions were observed, comprising 13 plant species and five ant species. The plant species included eight families, of which the Fagaceae family was the most represented (N = 5 species). For the ants, the Formicinae and Myrmicinae subfamilies were recorded, the latter being the most represented (N = 3 species) (Lara et al., 2020).

Given the difficulty of identifying some of the ant species in the field, Formica spp. Linneaus refers to the three morphospecies of the genus that are ecologically and functionally similar (Castillo-Guevara et al., 2019). Two of these species were later determined to be F. propatula Francoeur and F. pacifica Francoeur. Both ant species were found in the oak forest and grassland of the MNP. The ant-plant network of MNP’s oak forest is significantly nested (Lara et al., 2020). Five plant species (39% of the total registered species) and one ant species (Formica spp., 20% of the total recorded species) were identified as forming the network’s core.

In grassland, 33 interactions were observed, comprising 28 plant species and four ant species (Lara et al., 2020). The plant species included 17 families, of which Asteraceae was the most represented (N = 8 species) (Lara et al., 2020). For the ants, the Formicinae and Myrmicinae subfamilies were recorded, the latter being the most represented (N = 3 species) (Lara et al., 2020). The ant-plant network was not significantly nested (Lara et al., 2020). Three plant species (11% of the total registered species) and one ant species (Formica spp. 25% of the total recorded ants) were identified as forming the core of the network.

Ant-plant network FBSP

In oak forest, 55 interactions were observed, comprising 34 plant species and nine ant species (Fig. 1A, Data S2). The plant species comprised 19 families of which Asteraceae was the most represented (N = 8 species). For the ants, four subfamilies were recorded, and the most represented was Myrmicinae (N = 5 species) (see Table S2). A significantly nested pattern was observed in the ant-plant network (NODF = 43.13, NODF(Er) = 19.56, P < 0.01). Four plant species (12% of total recorded species) and one ant species (Prenolepis imparis Say, 11% of total recorded species) shape the core of the network (Fig. 1A).

In grassland, 63 interactions were observed, comprising 28 plant species and 12 ant species (Fig. 1B, Data S2). The plant species comprised 14 families of which Asteraceae and Agavaceae were the most represented (N = 6 species each). For the ants, four subfamilies were recorded, and the most represented was Myrmicinae (N = 6 species) (see Table S2). A significantly nested pattern was observed in the ant-plant network (NODF = 57.91, NODF(Er) = 20.75, P < 0.01). Four plant species (14% of total recorded species) and two ant species (Camponotus rubrithorax and Pseudomyrmex sp. 17% of total recorded species) shape the core of the network (Fig. 1B).

Ant abundance MNP

Only 200 of the 240 pitfall traps placed in the MNP were effective (i.e., 40 were damaged by animals or people). In total, 18 morphospecies (13 species and five genera) were registered, of which Formica spp. had the highest abundance in the oak forest (44 occurrences, i.e., number of times the ant was recorded) and the grassland (97 occurrences) (Table 1). One species was found exclusively in the oak forest (Lasius latipes Walsh), while three morphospecies were exclusive to the grassland (Dorymyrmex sp. Mayr, Brachymyrmex sp. Mayr, and Acanthomyops sp. Mayr). Fourteen of the registered morphospecies were shared by both types of vegetation (Formica spp., Temnothorax puctithorax Mackay, W.P., Temnothorax neomexicanus Wheeler, W.M., Temnothorax texanus Wheeler, W.M., Lasius niger Linnaeus, Pheidole soritis Wheeler, W. M., Pheidole chalca Wheeler, W.M., Pheidole sp. 1 Westwood, Myrmecocystus melanoticus Wheeler, W.M., Crematogaster lineolate Say, Monomorium minimum Buckley, Myrmica Mexicana Wheeler, W. A., Camponotus picipes pilosulus Emery and Solenopsis picea Emery)

Ant abundance FBSP

In the FBSP, only 226 of the 240 pitfall traps that were placed were effective. In total, 25 morphospecies (10 species and 15 genera) were registered. In the oak forest, Prenolepis imparis had the highest abundance (71 occurrences), and in the grassland was Camponotus rubrithorax (41 occurrences) (Table 1). Twenty-five morphospecies were recorded, of which seven were exclusive to the oak forest (Prenolepis imparis, Neivamyrmex sp. 1 Borgmeier, Myrmica mexicana, Pheidole sp. 5, Temnothorax andrei Emery, Stenamma manni Wheeler, W. M. and Camponotus sp. 3 Mayr), and seven exclusive to the grassland (Camponotus rubrithorax, Hypoponera sp. Santschi, Pheidole tolteca Forel, Dorymyrmex insanus Buckley, Pheidole sp. 6, Temnothorax augusti Baroni Urbani and Neivamyrmex sumichrasti Norton) and 11 in both vegetation types (Pheidole tepicana Pergande, Pheidole hirtula Forel, Linepithema dispertitum Forel, Pheidole nubicola Wilson, Temnothorax andrei, Monomorium ebenium Forel, Crematogaster sp. Lund, Temnothorax brevispinosus Mackay W.P., Camponotus sp. 2, Solenopsis pollux Forel and Labidus coecus Latreille).

Dominance hierarchy MNP

For the dominance hierarchy experiment in the MNP, 648 baits were placed. Ants were recorded in 347 baits (53.5%), none were recorded in 263 (40.6%), and 38 baits (5.9%) were damaged by local fauna. Nine species of ants were recorded in the baits (Formica spp., T. punctithorax, T. texanus Wheeler, W.M., Camponotus spp., M. mexicana, Monomorium minimum, Pheidole soritis, Pheidole sp. 2, and Lasius niger). In the oak forest, six ant species were registered (Formica spp., T. punctithorax, T. texanus, M. mexicana, M. minimum, P. soritis), of which only three ant species were present in the ant-plant network. In the grassland, eight ant species were registered (Formica spp., T. punctithorax, Camponotus spp., M. mexicana, M. minimum, P. soritis, Pheidole sp. 2, and L. niger), of which only three ant species were present in the ant-plant network. Likewise, although Crematogaster lineolata, formed part of the network, it was not found at the baits.

The DI was determined only for the three ant species found in the baits and the oak forest network, i.e., Formica spp., T. punctithorax, and M. mexicana. The DI indicated that the behavior of all ant species was submissive (DI <0.5) (Table 1). Also, in the grassland, the DI was determined only for the three ant species found in the baits and the network, i.e., Formica spp., T. punctithorax, and M. minimum. The DI indicated that the behavior of all ant species was submissive (DI < 0.5). M. minimum was the least submissive of the grassland species, with a value of 0.31 (Table 1). In the oak forest, the DI of the core species in the network (Formica spp.) was not significantly different of the species at the periphery (T. punctithorax and M. mexicana) (DI Core: 0.04 ± 0.08, DI Periphery: 0.35 ± 0.50, t =  − 1.520, P = 0.1595). In the grassland, the DI of the core species in the network (Formica spp.) was significantly higher than of the species at the periphery (T. punctithorax and M. minimum) (DI Core: 0.06 ± 0.05, DI Periphery: 0 ± 0, t = 4.001, df = 13, P = 0.0015).

Dominance hierarchy FBSP

In the FBSP, ants were only registered in 170 out of the 504 baits (33.73%). Thirteen ant species were recorded on these baits (M. ebenium, P. imparis, L. dispertitum, P. tepicana, C. rubrithorax, P. nubicola, D. insanus, T. andrei, T. rugithorax Mackay, W.P., T. augusti, T. brevispinosus, Nylanderia austroccidua Trager and Pheidole hirtula). In the oak forest, of eight species registered at the baits (M. ebenium, P. imparis, L. dispertitum, P. tepicana, T. andrei, T. rugithorax, T. brevispinosus, and P. hirtula), only four species were registered in the ant-plant network (P. imparis, L. dispertitum, P. hirtula, and M. ebenium, Fig. 1A, Table 1). While in the grassland, of nine species registered at baits (M. ebenium, L. dispertitum, C. rubrithorax, P. nubicola, P. tepicana, P. hirtula, D. insanus, T. augusti, and N. austroccidua), only seven species were present in the ant-plant network (Fig. 1B, Table 1).

The DI was determined only for ants that were found both at the baits and in the network, i.e., in the oak forest, for P. imparis, L. dispertitum, P. hirtula, and M. ebenium; in the grassland, for C. rubrithorax, M. ebenium, D. insanus, P. tepicana, P. nubicola, P. hirtula, and N. austroccidua (Table 1). The DI indicated that all ant species behave as submissives (DI <0.5), with M. ebenium being the least submissive, with a value of 0.39 in the oak forest, and D. insanus with a value of 0.33 in the grassland (Table 1). In the oak forest, when comparing the DI of the core species of the network (P. imparis) with those of the periphery (L. dispertitum, P. hirtula, and M. ebenium), no significant differences were found (DI Core: 0.280 ± 0.208, DI Periphery: 0.245 ± 0.323, t = 0.259, df = 19, P = 0.7982). In the grassland, when comparing the DI of the core species of the network (C. rubrithorax) with those of the periphery (M. ebenium, D. insanus, P. tepicana, P. nubicola, P. hirtula, and N. austroccidua), no significant differences were found (Z =  − 0.573, P = 0.5914).

Modified dominance index MNP

In the oak forest of the MNP, the average MDI of the species was different from the DI, indicating that Formica spp. was the most dominant species (MDI = 0.78), and the behavior of the other species was submissive (MDI < 0.5) (Table 1). In the grassland, Formica spp. was also the most dominant species (MDI = 0.61), and the behavior of the other species was submissive (MDI <0.5). However, in the oak forest, no significant differences were found between the MDI of the core species of the network (Formica spp.) and those of the periphery (T. punctithorax and M. minimum) (MDI Core: 0.77 ± 0.081, MDI Periphery: 0.412 ± 0.472, t = 1.868, df = 10, P = 0.0913). Also, in the grassland, no significant differences were found between the MDI of the core species of the network (Formica spp.) and those of the periphery (T. punctithorax and M. mexicana) (MDI Core: 0.606 ± 0.138, MDI Periphery: 0.384 ± 0.374, t = 1.377, df = 13, P = 0.1918).

Modified dominance index FBSP

In FBSP, the average MDI of the species differed from the DI. In oak forest, M. ebenium was the species with the highest dominance value (MDI = 0.74), followed by L. dispertitum and P. imparis (which were also dominant with MDI values >0.5). Finally, P. hirtula had values <0.5, which classifies them as submissive species (Table 1). In the grassland, P. nubicola was the species with the highest dominance value (MDI = 1), followed by C. rubrithorax, P. tepicana, D. insanus P. hirtula, and M. ebenium. (which were also dominant with MDI values >0.5), and finally, N. austroccidua had values <0.5, which classifies them as a submissive species (Table 1). However, in the oak forest, no significant differences were found between the MDI of the core species (P. imparis) and those of the periphery (L. dispertitum, P. hirtula, and M. ebenium) of the network (MDI Core: 0.600 ± 0.36, MDI Periphery: 0.571 ± 0.42, t = 0.152, df = 19, P = 0.8809). Also, in the grassland, no significant differences were found between the MDI of the core species (C. rubrithorax) and those of the periphery (M. ebenium, D. insanus, P. tepicana, P. nubicola, P. hirtula, and N. austroccidua) of the network (MDI Core: 0.833 ± 0.16, MDI Periphery: 0.58 ± 0.39, t = 1.554, df = 22, P = 0.1345).