Introduction

Land use change is a major driver of biodiversity depletion, mainly because of its devastating effect on habitat availability and the consequent extinction of species (Sala et al. 2000; Pimm et al. 2014; Ceballos et al. 2017). These changes in natural landscapes are strongly linked to the replacement of mature forest areas for intensive agroforestry, agriculture, and livestock farming (Hansen et al. 2013; Vale et al. 2019). Between these impacts, the conversions of forest to livestock farming is one of the major causes of deforestation in the Brazilian Amazon, one of the most diversity areas of the planet (e.g., Barona et al. 2010).

The colonization of beef cattle has played an important role in Brazil’s economy (Walker et al. 2013; ABIEC 2019). Brazil is currently the largest producer and exporter of commercial livestock in the world (Meyer and Rodrigues 2014; ABIEC 2019). This brought negative implications to Amazon forest (Vale et al. 2019), because livestock farming tends to increase deforestation (Walker et al. 2013; Fearnside 2017). Between 2004 and 2019, the Brazilian Amazon was deforested at an average rate of 10,021 km2 per year (INPE 2020), and the majority forest area lost in the Brazilian Amazon eventually became pastures (Barona et al. 2010). Therefore, understanding the land-use change and its effects on biodiversity, especially in the Brazilian Amazon, is a critical starting point to make decisions about effective conservation strategies (Phillips et al. 2017).

Ants (Hymenoptera: Formicidae) are a taxonomic group with a long history as biological indicators (Alonso 2000; Andersen et al. 2002; Gerlach et al. 2013; Carvalho et al. 2020). This dominant group is present in most Neotropical and subtropical habitats of native forests up to pastures and farmlands (Carvalho et al. 2020). Overall, some ant species are highly resilient to habitat disturbance (Anjos et al. 2017; Andersen 2018), however few studies have shown the responses of Amazonian ant communities to human disturbances (see Vasconcelos et al. 2006, 2010), and across different land uses in the landscape (Solar et al. 2016). In response to habitat disturbances, the abundance of dominant species mediated by changes in vegetation structure can induce changes in richness and composition of ant assemblages (Solar et al. 2016) showing a direct—and negative impact—of deforestation on ant diversity (Yates and Andrew 2011; Kuate et al. 2015).

In this study, we directly studied the diversity of ground-dwelling ant species in secondary forests and neighboring pastures areas. These two landscape components are becoming common habitats in the Amazon (Barona et al. 2010; Fearnside 2017), specially for insects, e.g., ants. Ground-dwelling ants play crucial roles in the ecosystems such as natural processes (e.g. nutrient cycling, soil aeration) (Folgarait 1998) and forest regeneration (e.g. seed dispersal) (Gallegos et al. 2014; Anjos et al. 2020). In addition, this ant fauna is diversity and sensitivity to environmental changes (Ribas et al. 2012), which makes this group suitable to be used on studies that aim to compare diversity between habitats. For this reason, we evaluated the species richness and composition of ant communities occurring in secondary forests and in pastures of the Amazon. We hypothesized that the conversion of secondary forest to pasture: (i) impoverish the ant species diversity and (ii) promote changes in ant assemblage composition. In this way, specific ant species are expected to be defined for each kind of environment.

Materials and methods

Study site

The fieldwork was carried out in the municipality of Autazes, 112 km from Manaus, Amazonas state, Brazil (Fig. 1). Autazes has an economic-based in buffalo and livestock farming, being among the ten largest producers of cattle of the state, and it also appears in the ranking of the 10 municipalities that more contributed to deforestation in 2008 and 2009 in the Amazonas state (Piontekowski et al. 2011). Between 2008 and 2019 the clearing rate at Autazes was 190.91 km2 and the municipality occupies the 111th position of accumulated deforestation increments in the Brazilian Legal Amazon (Assis et al. 2019; INPE 2020). According to Köppen’s classification, the climate is tropical humid with a mean annual temperature 26 °C and a mean annual precipitation of 2200 mm, which is distributed seasonally (Ribeiro 1976). Thus, the weather is similar to Manaus, where the precipitation varies seasonally, raining between November and May and the dry season is between June and October.

Fig. 1
figure 1

Map of the study area in Autazes, Amazonas, Brazil

Full size image

The samplings were done between 6 and 21 September of 2013 in areas that had been flooded in the past (e.g., Galvão et al. 2011). We selected five areas for ant sampling for this study [Vista Alegre = VA; Soares (Comunidade) = SO; Soares (Seu Olavo) = OL; Soares (Tiririca) = TI; Juauri = JU] (Fig. 1; Appendix A01). The regional biome is Amazon rainforest. Secondary forests are characterized by a set of small forest fragments, poorly preserved, immersed in a highly fragmented landscape. These forests suffered from the loss of large trees, increased invasion of exotic species and the intensive hunting of wild animals (Chokkalingam and De Jong 2001). While pastures are characterized by open areas, where forests have been cleared and replaced by grasses to feed livestock, at least since the 1990s (Appendix A01).

Ant sampling

For each habitat, we selected three sites of native Amazon forest that have been replaced by pasture for livestock and three sites in adjacent secondary forests (total = 30 ground transects), distanced 200 m apart between pasture and forest transects (Fig. 1). In each site were placed 10 pitfall traps (Bestelmeyer et al. 2000) on the ground spaced to 20 m along a transect. Thus, a total of 150 traps were installed for the three sites of forests and for the three sites of pastures, in total 300 traps. The pitfall traps were composed of disposable plastic cups with 150 ml volume, 6.5 cm diameter and 7 cm in depth, filled with a solution of water with few drops of soap (to break the superficial tension of water) (Support information, Appendix A02). The pitfalls were exposed for 48 h. The ants collected were identified at the higher taxonomic level possible using of key from Baccaro et al. (2015) and by comparison with the reference collection of the “Laboratório de Mirmecologia, Centro de Pesquisas do Cacau” (CPDC collection, CEPEC/CEPLAC) where vouchers were deposited. The nomenclature follows AntWeb (2020).

Data analysis

First, we performed a Mantel test (Primer package) with a weighted Spearman rank correlation to address the effect of plot spatial location on levels of taxonomic similarity in PRIMER v6. Straight-line distances between plots were ln-transformed, as suggested by Condit et al. (2002). The Mantel test was carried out considering a group of 14 forest and 14 pasture areas, because two transects (one for forest and one for pasture) did not return any ant, so both were excluded both from the analysis. A Mantel test failed to uncover any spatial effects on the taxonomic similarity among plots (Rho = 0.063; p = 0.133).

Thus, in total, 14 forest plots and 14 pasture plots were analyzed. The occurrence of ant species per pitfall trap was used as a measure of the species' abundance. Differences in average species richness between the two habitat types were analyzed using t test using PAST (Hammer et al. 2001). To compare the overall ant species richness at habitat spatial scale we generated species-individual accumulation curves. Species accumulation curves permit comparing the diversity of ant assemblages at comparable levels of sampling effort. Species accumulation curves of the observed numbers of species (Sobs) were computed for the ant species in the two environments studied. Computing 95% confidence intervals (Colwell et al. 2004) allows for a statistical comparison of the species richness of the two datasets (Colwell et al. 2004). Species accumulation curves were computed with the analytical procedure in the EstimateS 9, using 500 randomizations and sampling without replacement (Colwell 2013).

To analyzed changes in the ant species composition we considered the patterns of beta diversity across the landscape. For this purpose, we calculated the average number of species present per site defined as β = γ – α, where γ is the number of species sampled in the entire landscape (gamma diversity) and α is the average number of species present at a given site (alpha diversity) (Veech et al. 2002) using EstimateS 9. This approach is used as a measure of additive partitioning of diversity and allows a direct comparison between alpha and beta diversities in terms of numbers (or percentage) of species.

Nonmetric Multidimensional Scaling (NMS) based on Sorensen’s distance were performed to whether there are community similarities between areas (Kruskal 1964; Mather 1976) using the PC-ORD software package (McCune and Mefford 1999). The data underwent the transformations and relativizations option in PC-ORD. NMS was constrained to two axes and the “slow-and-thorough” autopilot mode of NMS in PC-ORD used the best of 250 runs with a random starting configuration using the real data and 250 runs using randomized data for a Monte Carlo test of significance. Final stability was evaluated for each run by examining plots of stress (a measure of the dissimilarity between ordinations in the original p-dimensional space and the reduced dimensional space) versus the number of iterations. We overlaid regional categories on the final plot, and used joint plots, to examine how habitat types (forest and pastures) on to the ordination axes. The length and direction of a joint plot is based on the strength of the correlation between a variable and the ordination axes (McCune and Grace 2002). Variables with the strongest patterns of correlation with the ordination structure, based on Pearson R2 values > 0.20 with one or more of the axes. Finally, we performed an Indicator Value (IndVal) analysis (Dufrêne and Legendre 1997) performed for each species in relation to their distribution in both habitats in PC-ORD. We used randomizations to determine the statistical significance of IndVal (Monte Carlo test; p < 0.05), therefore, ant species with high specificity within a habitat type will have a high IndVal. The position of top indicator species was identified in rank abundance plots.

Results

A total of 46 ant species from 23 genera were recorded (Support information, Appendix A03). For secondary forest areas, 36 ant species from 20 genera were recorded, being 23 exclusives species. The richest genera were Pheidole (5 spp.), Solenopsis (4 spp.) and Ectatomma (4 spp.). Whereas in pasture, we recorded a total of 23 ant species from 13 genera, being 10 exclusive species. The richest genera were Pheidole (5 spp.), Solenopsis (3 spp.), and Camponotus (3 spp.). Forests and pastures shared 13 ant species (~ 28% of total species). Besides, the ant rarefaction curve in forest had 1.4-fold more species as compared to neighboring pastures (Fig. 2a). However, despite the apparent difference in absolute richness, the results showed that average species richness did not differ statistically between habitats (t = 0.13; df = 26; p = 0.89).

Fig. 2
figure 2

a Individual-based rarefaction curves for ground-dwelling ants in pasture and forest areas in the Amazon forest, north Brazil. Bars dotted indicate 95% confidence intervals. b Individual-based rarefaction curves for ground-dwelling ants in the Amazon forest, north Brazil. Showing an additive diversity form as β = γ – α

Full size image

The gamma diversity can be expressed in an additive form as: 46 [γ] = 39.7[α] + 6.3[β] (Fig. 2b). Beta diversity contributed 13.9 and 13.5% of the total gamma diversity for forest areas and pasture areas, respectively. The NMS analyses based on species similarity resulted in two segregated clusters: one formed by forest, one by pasture (stress = 0.16). Joint plots indicated a correlation with the axis 2, suggesting that the habitat type was driving the separation taxonomic by region in the ordination (r = 0.66, p = 0.05, Fig. 3). Finally, an indicator species analysis underscored the occurrence of three species (Ectatomma lugens, Neoponera apicalis, and Pachycondyla crassinoda) that can be considered typical of forest areas, while a subset of four species (Camponotus blandus, C. melanoticus, Ectatomma brunneum, and Odontomachus haematodus) stands out from pasture areas (Table 1). The four indicator species with higher IndVal in pasture areas add 63% to all species distribution (Fig. 4a), indicating a dominance of species with high abundance. In contrast, the three indicator species in forest areas only add 34% to the whole distribution, indicating that the ant assemblages in this habitat have relatively a high richness of species with low abundance (Fig. 4b).

Fig. 3
figure 3

Nonmetric multidimensional (NMS) ordination of ground-dwelling ant community across pasture (white circles) and forest (black circles) areas based on the abundance data in the Amazon forest, North Brazil

Full size image
Table 1 The Indicator Value (IndVal) of ground-dwelling ant species in forest and pasture areas in the Amazon forest, north Brazil
Full size table
Fig. 4
figure 4

Rank distribution by abundance of ground-dwelling ant species sampled using pitfall traps in a forest remnants (black circles) and b adjacent pastures (white circles) in the Amazon forest, north Brazil. The arrows show the identity and position of top indicator species within of species-rank curves (see Table 1)

Full size image

Discussion

This study presented the effect of conversion from secondary forests to pastures on ground-dwelling ant communities in central Amazonia. Although the results have not shown a reduction in ant richness due to the land use change by livestock farming, our results showed that the taxonomic composition varies between habitat types. The substitution of native forest for pastures has been implemented in many regions of the Brazilian Amazon, with the aim of improving productivity of livestock (Vale et al. 2019). Changes in vegetation structure can directly influence the composition of the ant fauna, due to other variables, such as luminosity, temperature, humidity (Gibb et al. 2015), and aboveground biomass (Solar et al. 2016) that directly and indirectly affect the foraging, prey availability and nesting sites to the ant species. Similar results were also found by Vasconcelos (1999) that found a distinct ant species composition between pasture and mature forest.

The similarity of ant species richness between forest and pasture areas can be explained due to some forest areas can serve as sources of species for disturbed areas, corroborating the landscape species pool hypothesis (Tscharntke et al. 2012). We have to consider that a short distance between areas can facilitate these species flow. For instance, Frizzo and Vasconcelos (2013) shows that pasture areas in Brazil supports a diversity of species of growd-dwelling and arboreal ants similar to preserved areas of tropical savanna. Nevertheless, the ant species richness was low in both areas, although most studies have used pitfall traps associated with attractive baits (e.g., using sardine and honey), a different method of collection (see Vasconcelos 1999; Solar et al. 2016). This low overall ant richness can be useful to define the conservation status of secondary forest areas, in the Amazon or other ecosystems. On the other hand, we found a trend for forest areas to present higher species richness than disturbed ones, as well in other Amazonian forests (Vasconcelos 1999; Solar et al. 2016). For instance, Menezes and Schmidt (2020), in the southwestern Brazilian Amazon using five types of liquid food resources to sampling ants showed higher ant species richness in forest than in pasture areas. These authors, in total (considering both forest and pasture areas), found more than twice ant species compared with our results, but recorded almost a half of ant species in pastures areas. Overall, areas that have been used as pastures appear to be recovering more slowly than other disturbed areas (Vasconcelos 1999). Therefore, in addition to local environmental differences and the sampling methods used, these dichotomous responses confirm the limited value of species richness to access the effects of converting forest areas to pastures, while species composition may better fulfill this role (Barlow et al. 2007; Solar et al. 2016).

Ant community composition was different between secondary forests and pastures, which we can assume that 60% of the taxonomic composition was driven by the type of habitat. These results are consistent with previous studies in other areas of the Amazon forest (Vasconcelos 1999; Wilkie et al. 2009; Solar et al. 2016). Differences in species composition may reflect in the ecosystem services provided by them (Del Toro et al. 2012). For instance, two of three indicator species of secondary forests are specialist predators which are more sensitive to environmental disturbances (e.g., species of Pachycondyla, and Neoponera genera) (Leal et al. 2012; Brandão et al., 2015). Moreover, another indicator species, Ectatomma lugens, also is a specialist predator very abundant in pristine Amazon forests (Vasconcelos et al. 2000). As a result, areas of secondary forests, through specialist predator ants, can maintain more ecosystem services such as the top-down effects on control of small invertebrate populations and seed dispersal, for instance (Passos and Oliveira 2002; Leal et al. 2012). Therefore, even with low ant species richness and an intermediate level of degradation, secondary forest areas can provide important ecosystem services mediated by their ant community reinforcing the importance of these areas in the landscape.

In pasture areas, Camponotus species represented half of the indicator species. This genus consists of a high diversity of generalist ants common in open and disturbed habitats (Anjos et al. 2017; Alves-Silva et al. 2020). Despite the predatory habit, Ectatomma brunneum is also abundant in open fields or degraded areas (Gomes et al. 2009). An unexpected result was Odomomachus haematodus as an indicator species of pasture areas. Despite O. haematodus species belonging to a genus (Odontomachus) of predator species sometimes specialized (Brandão et al. 2015; Silva et al. 2015), this species seems to be very abundant in both forest and fragments of Amazonian forests. For instance, this species was one of the few commonly sampled in all 34 areas studied by Vasconcelos et al. (2006) in Amazon. Therefore, the occurrence of O. haematodus as an indicator species can be justified since the IndVal is strictly based on the relative abundance of the species (Dufrêne and Legendre 1997). Finally, beta diversity is reported to be low for ants in the Amazon (Vasconcelos and Vilhena 2006). Our study showed that the environment types share 63% of ant richness species. Previous studies also found low beta diversity and consequently a high percentage of shared species, being that 64% of ant species in less species-rich habitat (e.g., savanna) were also found in richer ones (e.g., forests) (Vasconcelos and Vilhena 2006), both habitats without anthropization, and 85% of the ant species found in fragments were also found in areas of continuous forest (Vasconcelos et al. 2006).

Conclusions

Overall, the results from our study support the alterations in ant diversity caused by livestock farming in the Amazon forest (Vasconcelos 1999; Solar et al. 2016). Despite the absence of changes in species richness, secondary forest areas and pastures showed a clear difference in assemblage composition. The next steps are to identify which factors explain the change in species composition between secondary forests and pastures.