Distribution and diversity of leaf-cutting ants in Northeastern Argentina: species most associated with forest plantations

Abstract Leaf-cutting ants (LCA) are considered one of the main herbivores and one of the most destructive pest insects of the Neotropics. Northeastern Argentina harbors the greatest species richness of these ants and in turn comprises the highest surface with forest plantations. Our aim was to establish which species of leaf-cutting ants are most commonly associated with forest plantations by analyzing their geographic distribution using published and unpublished species occurrence data. Also, estimate their potential areas of distribution along a latitudinal gradient that entirely encompasses northeastern Argentina using Ecological Niche Modeling. Only seven of the 20 species recorded were strongly associated with productive systems along the gradient, but only 2–3 species in each region could be considered high-risk species for forest plantations. High-risk species composition shows a turnover between regions. Our models show the potential distribution areas where LCA could become more abundant and dominant, and possibly causing a detrimental effect on the forest plantations in the studied region. We find that ecological niche models are useful tools to assess the environmental suitability of important LCA.


Introduction
The potentially invaded areas of a species depend on its dispersion capabilities and/or the possibilities of access and establish in new areas. It has been shown that pest species of leaf-cutting ants (LCA) are favored by the existence or creation of open areas and forest edges, roads and urban areas, such as gardens and parks (Urbas et al. 2007). For example, a study in Brazil found that Atta cephalotes and Atta sexdens can increase their nest densities with the creation of forest edges . In this context, the growing extension of forest plantations and crops throughout the northeast of Argentina could play an important role in the current distribution and potential expansion of LCA. Land use in northeastern Argentina has changed considerably in recent decades, where marginal agricultural lands and native forests were replaced by exotic tree plantations (Grau and Aide 2008). Natural forest cover decreased by 49%, for example, in the province of Misiones between 1973 and 2006, whereas the area of commercial forest plantations increased by 10% (Izquierdo et al. 2008). Similarly, around 60% of the native forest was lost and/or replaced by productive systems in Corrientes province in a shorter period (1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001) (Guida Johnson and Zuleta 2013). The replacement of native forests by a monoculture of an exotic commercial forest species has likely decreased the local LCA diversity by favoring some few and more ecologically dominant LCA species. However, it is difficult to predict how the distribution and diversity patterns of these ants are affected under this land use change, and therefore, to establish which will be the most successful on a local and regional scale, as their native ecosystem and environmental conditions change.
These ants belong to the genera Atta and Acromyrmex, and as fungus-growing ants, they can be characterized by the habit of feeding on a basidiomycete fungus cultivated by collecting and preparing an adequate substrate of leaf and flowers for the development of fungus (H€ olldobler and Wilson 2011). They are exclusively Neotropical and can be found from southern United States to Argentina. Most LCA act as key components of ecosystems due to their role in nutrient cycling and regulation of primary production Farji-Brener and Werenkraut 2015), and have also been regarded as ecosystem engineers, due to their availability to modulate resources for other species by causing physical changes in biotic or abiotic materials (Montoya-Lerma et al. 2012;Meyer et al. 2013).
Although only a few LCA species are considered important pests (Della-L ucia 2003) those that occur in cropping systems are highly relevant pests. It has been estimated that the economic losses may reach a billion-dollar each year (Cherrett 1986;Montoya-Lerma et al. 2012). In Argentina, damage caused by these ants has been mentioned, and recognized as pest, since before beginning of the past century (Kusnezov 1963;Bonetto 1959). The systems most affected by defoliation of LCA are forest plantations and vineyards. A single LCA nest has a high requirement of plant material for the maintenance of thousands of workers (P erez 2009). To satisfy this demand, they must cut a vast amount of leaves and flowers, which in turn can affect the growth and survival rate of plantations (Hern andez and Jaff e 1995). Depending on the density and size of their nests, the age of the plantation and the environmental conditions at the time of the damage, LCA can eliminate important sectors of young plantations in a short time (FAO 2006). The northeast of Argentina (provinces of Misiones, Corrientes, Entre R ıos and Buenos Aires) comprise around 80% of the forest plantations in the country (MAGyP 2012); the most cultivated species are Pinus spp., Eucalyptus spp. and Salicaceae (Salix and Populus spp.). It is also the same region where the highest richness of LCA is located (Kusnezov 1963; Mayh e-Nunes and Jaff e 1998).
The available information on the distribution of LCA in this region is scattered and the known geographic distribution ranges still have a coarse resolution. Determining the actual geographical distribution of LCA, their association with forestations and the environmental factors that determine their distribution will allow for making better forest management decisions in regions where LCA species richness is high. In this work, we first address which LCA species are distributed along a latitudinal gradient in northeastern Argentina, the region with the highest surface of commercial forest plantations. Using this as a baseline, we identify which species are most related to commercial forestations. Finally, we modeled the potential geographic distribution of the LCA species most associated to forestations in each region along the latitudinal gradient, in order to identify the putative areas that could have a considerable increase in abundance and distribution Figure 1. Leaf-cutting ant richness across latitudinal gradient from the low delta of the Paran a River in Buenos Aires, crossing the Argentine Mesopotamia, to the Paran a jungle in Misiones. Each cell (1 Â 1 degree) shows the number of species (richness) and blue dots the occurrence localities of the 20 LCA species found in this study. Upper left plot shows the accumulation and richness estimator curves. Bottom left plot shows how the richness numbers change along the latitudinal gradient, where the black dots are the number of species per one degree of latitude and the grey line is the moving average. ranges of these ants with the intensification and/or changes in the land use.

Study region
The study region covers the entire northeast of Argentina and encompasses from the low delta of the Paran a River in Buenos Aires province, crossing the Mesopotamian provinces of Entre R ıos and Corrientes, to finish in the Atlantic Forest in Misiones province. The study region contains the ecoregions of Delta and Paran a Islands, Pampa, Espinal, Esteros del Iber a, Campos and Malezales and Paranaense (Burkart et al. 1999). Forest plantations are usually established in open grassland habitats mostly in the Mesopotamian provinces, whereas native forests are typically replaced by commercial forestations in the highlands of the low delta of the Paran a River in Buenos Aires and Entre R ıos and the Atlantic Forest in Misiones (Grau and Aide 2008).

Data acquisition
Using data from surveys and records in published literature we built a matrix that relates the occurrence (latitude, longitude and altitude) of all LCA species found throughout the study region ( Figure  1). Field surveys were conducted in natural (wildlife) and anthropic (urban and plantations) areas across the study region. All surveys were made in different field trips made in Entre R ıos, Santa Fe, Corrientes and Misiones provinces. Each field trip consisted in directed search for LCA nests and the collection of its related data (habitat, coordinates and type of nest). A representative sample of each species was collected in each locality, preserved in 96% ethanol and morphologically identified at the laboratory. The published occurrence data used were obtained from peer-reviewed papers, technical reports, the GBIF (https://www.gbif.org) and AntWeb (https:// www.antweb.org) databases. The matrix obtained were validated and cleaned using Chapman's guidelines (Chapman 2005), location (e.g., municipality) names were updated to the most current names and classifications of administrative divisions, and records were geocoded by using GeoNames (http:// www.geonames.org) and Google Earth. Dubious occurrence records and the oldest records (previous to 1950) were excluded from the analysis because they could be reflecting occurrences that have already changed over time.

Species richness along the gradients
To examine patterns of species richness of LCA across northeastern Argentina, a grid of one by one degree of latitude and longitude was defined as a sampling unit and mapped in an extension that covers from northern Buenos Aires, most of Santa Fe, and all of Entre R ıos, Corrientes and Misiones provinces: the area where more than 80% of the commercial forest plantations are located in Argentina (MAGyP 2012). Within this grid were included all the confirmed occurrence points of LCA species inhabiting this region, plus records from neighboring countries up to a range of $100 km from the Argentine border. Using LCA occurrence and frequency data in each cell of the grid, a matrix was generated to estimate the representativeness of the sampling through an accumulation curve and the non-parametric richness estimators Jacknife 1 and Boostrap (calculated with the "Vegan" package in R; Oksanen et al. 2018).

Species associated with productive systems
The association of LCA within the different productive systems (forest plantations and crops) was established from the field records and by a review of the reports published in the literature for the study region. This allowed us to establish which of these species are the most commonly reported in these systems. The presence of LCA species in a particular system does not imply that it causes damage in it, therefore its pest status was defined mainly based on a consensus definition using the previously defined condition given in the literature (Fowler et al. 1989;Montoya-Lerma et al. 2012;Britto et al. 2016).

Potential species distribution
Only a few LCA species that reach locally high densities become important pests Nickele et al. 2009). To estimate the potentially suitable areas of geographical distribution of the most economically relevant LCA species according to their potential damage to forest plantations in northeastern Argentina, Ecological Niche Model (ENM) analyses were performed using a maximum entropy algorithm. These models can provide a pathway that statistically links the distribution of a particular LCA species with the spatial variation in the bioclimatic variables. In other words, the ENM allows obtaining correlative models that identify the most suitable area of a given species using the occurrence records of the species and a set of biologically relevant environmental variables. Also, ENM is a useful tool to test hypotheses about Table 1. Detrimental leaf-cutting ants with confirmed presence in forest plantations, vineyards and grassland systems throughout northeastern Argentina (including close border regions), based on literature records and own data. species range characteristics, niche partitioning or niche conservatism (Franklin and Miller 2010). The conceptual framework we assume is that described by Sober on and Nakamura (2009), where an ENM means to locate a species, given a physical space that meets the environmentally favorable areas for the species; in other words, to model the potential geographic distribution areas (G I or A in the BAM diagram; Sober on and Nakamura 2009). The models were only made for the LCA species that we found mostly associated with productive systems (Table 1) and that also had a high pest status according to the status given by Fowler et al. (1989), Montoya-Lerma et al. (2012) and Britto et al. (2016). To generate the models, we considered the complete distribution for each of the selected species, and therefore, we included records from their entire distribution range. The modeled species were: Acromyrmex ambiguus, Ac. crassispinus, Ac. heyeri, Ac. lobicornis, Ac. lundii, Atta sexdens and At. vollenweideri. These records were previously cleaned of duplicates and thinned in order to reduce the effect of uneven or biased occurrence records that can affect the performance of the models (Kramer-Schadt et al. 2013). We reduced the dataset with a distance criterion using the R package "spThin" (details in supplementary material; Aiello-Lammens et al. 2015). Twenty environmental variables were considered (19 bioclimatic, plus elevation). Bioclimatic variables were downloaded from Worldclim version 1.4 (Hijmans et al. 2005; http://www.worldclim.org/). These variables, commonly used in ENM and species distribution modeling, have been shown to be biologically meaningful to define the eco-physiological tolerances of the species (Graham and Hijmans 2006). Of these variables, a final set of ten was selected using a test for multicollinearity by examining cross-correlations (Pearson's correlation coefficient) and the variance inflation factors (VIFs) using the R base and "usdm" package (details in supplementary material; Naimi et al. 2014). All models were generated using MaxEnt version 3.4.1 (Phillips et al. 2006). This software uses the data of the sites where the species was recorded together with the environmental conditions, and searches for those distributions that comply with a series of environmental constraints, finding the areas with the most similar conditions to the sites where the species occurs (Phillips et al. 2006), in other words, the distribution probability of maximum entropy. The models were generated by choosing the logistic output, a regularization multiplier of 1, 10 replicates of cross validation tests, 10,000 background points (as pseudoabsences) and the functions of environmental variables set as default. To evaluate the performance of the models the threshold-independent area under curve (AUC) of the receiver operating characteristic (ROC) (Swets 1988) was calculated by randomly splitting the locality data into 75% for training and 25% for testing. This test measured the predictive accuracy of a model to distinguish between presence records from absence or background points (Phillips et al. 2006). Each model was plotted discriminating the environmental suitability of the MaxEnt output in four levels (<0.2; 0.2-0.4; 0.4-0.6; >0.6). Complementary plots where then generated by overlapping these values with the forest plantations map of Argentina (Ministerio de Agroindustria 2013) in order to show how these occurrence probability values are distributed in relation to the forest cover for each species. Finally, to quantify the overlap between the models, we calculated the D of Schoener (1968) & I indices proposed by Warren et al. (2008), by using the R package "ENMTools" (Warren et al. 2017). These indices can take values between 0 and 1, indicating overlap, absence or total overlap when the models are completely different or identical, respectively.

Study richness along the gradient
We recorded a total of 360 validated occurrence records (320 of Acromyrmex spp. and 40 of Atta spp.) belonging to 20 species of LCA distributed along the entire northeast of Argentina. The richness estimators indicated a good representation of the LCA species that inhabit northeastern Argentina, about 83-92% of the species were estimated to occur in this region of the country ( Figure  1). The highest richness of LCA was found in the north of Misiones ($12 spp.), whereas the lowest richness was in the low delta of the Paran a River region ($3 spp.), showing a strong decrease in the species richness with the increase in latitude ( Figure  1). The center-west region of Santa Fe province also showed a significant number of LCA species ($8 spp.), in contrast to areas in the southern-west of the study region, like Entre R ıos and Corrientes provinces, where only 3 to 4 species were found.

Species associated with productive systems
Of these 20 species, 14 were reported as present within forest plantations and other similar productive systems (e.g. corn crops), but only seven were strongly associated with them along the entire gradient (Table 1). Despite the decrease in the species richness associated with latitude, the number of high-risk species for forestations was relatively equal along all the latitudinal gradient (2-3 spp.), but there was a change in the composition of these 2-3 species (a replacement) throughout the different regions of the gradient. Acromyrmex lundii and Ac. heyeri were the species with most reports in different forest plantations, followed by Ac. subterraneus. However, most records for the latter species are only from plantations in Brazil. Forest plantations of Pinus taeda and Eucalyptus grandis presented a high diversity of LCA species (between nine and six species, respectively), whereas willow and poplar forest plantations had only two species (Ac. lundii and Ac. ambiguus). Here is important to consider that this latitudinal pattern may be a confounding factor, since Eucalyptus and pine plantations are more common in the central and northern part of the study area, while willows and poplars are produced mainly in the southern part of the study area. Reports on the presence of LCA in other systems were scarce and not very informative. For example, Anglada et al. (2013) reported the presence of LCA, especially Ac. lundii, in alfalfa (Medicago sativa), sunflower (Helianthus annuus) and corn (Zea mays).

Current and potential species distribution
The center of the geographical distribution of most of the LCA species found throughout northeastern Argentina is in this region or adjacent to it (e.g. the populations of Ac. lundii, Ac. heyeri, At. vollenweideri). In contrast, the populations of Ac. crassispinus, At. sexdens, Ac. niger and Ac. nigrosetosus present in northeastern Argentina correspond to the southernmost populations of these species (the limit of their geographic distribution ranges). Acromyrmex lundii and Ac. heyeri were more frequent in northeastern Argentina, especially in the provinces of Entre R ıos, Corrientes and part of Santa Fe. Their ranges reach the south of Brazil and Paraguay. Other species, like Ac. balzani and Ac. nigrosetosus presented a more restricted distribution, that were found mainly in Misiones province and were usually associated with natural ecosystems. Acromyrmex evenkul was the only species reported in a single locality (Iguaz u National Park, Misiones). The two species of Atta were At. sexdens, distributed mostly in northern Misiones, and At. vollenweideri, mostly distributed in Entre R ıos and Santa Fe provinces.
All ENM models ( Figure 2) were very consistent with the known distribution of LCA species (Farji-Brener and Ruggiero 1994) and had high (>0.9) AUC values (Table 2), indicating a much better performance than that expected by random. Only the model for At. sexdens had a moderate performance, probably due to the scarce number of occurrence points in our study region because this species has a much wider distribution range outside of Argentina. Despite the differences in the patterns and breadth of the distribution ranges of the analyzed species, the contribution of the variables to the models was similar. The temperature seasonality (Bio04) showed the highest contribution percentage in all models (close to or higher than 60%), making it the most influential abiotic factor in the distribution of the LCA. The variables that contributed more than 10% (Figure 3) were the precipitation in the warmest quarter (Bio18) and the mean temperature of the wettest quarter (Bio08), which showed similar contribution percentages in Ac. crassispinus, Ac. heyeri, Ac. lobicornis, Ac. lundii and At. vollenweideri. The variables that contributed most to Ac. ambiguus were precipitation of driest month (Bio14; 29.3%) and precipitation seasonality (Bio15; 15.0%), and to At. sexdens, they were precipitation of wettest month (Bio13; 25.2%) and precipitation of driest month (Bio14; 12.4%).
All the species showed high levels of occurrence probability (P > 0.69) in at least some parts of the study region (Figure 2). The areas that showed the highest suitability in the models were: (a) south of Brazil and Uruguay for Ac. ambiguus and Ac. crassispinus; (b) low and middle basin of Paran a River (Argentine Mesopotamia) and south of Brazil for Ac. heyeri and Ac. lundii; (c) Argentine monte ecoregion and arid regions of south of Brazil and Uruguay to Ac. lobicornis; (d) Atlantic forest for At. sexdens; (e) low and middle basin of Paran a River (Argentine Mesopotamia) and Santa Fe province for At. vollenweideri (Figure 2). Forest platations cover are mostly concentrated in the province of Misiones ( Figure 2O), where At. sexdens was the LCA with the highest occurrence probabilities in these forest plantations ( Figure 2L), following by Ac. crassispinus and Ac. ambiguus. Other important portion of forest plantations are found close to the Uruguay river, in Entre R ıos province, where species Ac. lundii ( Figure 2J) and Ac. heyeri ( Figure  2F) shows the highest values in the models for this forest plantations. The region of the delta of Paran a is another region with an important concentration of forest plantations, where the species that showed high probabilities in the models were Ac. ambiguus ( Figure 2B) and Ac. lundii ( Figure  2J). The highest I and D overlap indices (Figure 4) were for the models of Ac. ambiguus, Ac. crassispinus, Ac. lundii and Ac. heyeri, while the models of At. sexdens and At. vollenweideri show the lowest overlap indices.

Discussion
When a natural environment is modified in a monoculture, we can expect LCA pest species to become more abundant and dominant, causing a probable impact to forestry production in the study area. Our models show the potential distribution areas where this could happen in relation to the existing forest cover and environmental conditions in the northeast of Argentina. This work also shows the applicability of ecological niche models in the generation of geographical pest distribution information, providing useful tools for the integrated pest management of LCA.

Study richness along the gradient
To our knowledge, the northeast of Argentina harbors the highest richness of LCA species (20 spp.) currently known for a single region. Only taking into account species and not subspecies, other studies in nearby regions have found: 10 Acromyrmex species in 290 localities of Rio Grande do Sul State in Brazil (Loeck et al. 2003); 6 and 9 Acromyrmex species in Santa Catarina and Sao Pablo States in Brazil, respectively (Rando et al. 2005); 13 Acromyrmex and Atta species in eastern Paraguay (Fowler 1985). On a Neotropical scale, the richness found in the northeast of Argentina is very high compared to other regions farther north. For example, there are only 12 LCA species in Colombia and only around 5 in Central America (Maes and Mackay 1993).
As in other insect groups that show latitudinal diversity gradient (Condamine et al. 2012), the LCA species richness decreased as latitude increase. The latitudinal richness gradient observed in LCA species has already been described by Fowler (1983a) and Farji-Brener and Ruggiero (1994), who explained it as a combination of climatic and structural complexity of the habitat and the biological attributes of LCA species. The type and form of construction of nests play an important role in the distribution of LCA in cold and temperate environments, especially those of the genus Acromyrmex because they serve as a "thermal buffer" that helps to diminish the negative effect of the extremely low temperatures on the fungus (Farji-Brener 2000). In this sense, the soil temperature regimes have been found to be correlated with the nesting habits in several South America Acromyrmex species (Bollazzi et al. 2008), a relationship that can play an important adaptive role in the construction of superficial or subterranean nest, and thus to influence their distribution extent and range. Indeed, the LCA species that we found further south in the gradient tended to build nests with domes (like Ac. ambiguus, Ac. lobicornis, Ac. crassispinus) compared with more subtropical areas where a higher diversity of nest types can be found, such as nests forming soil mounds or completely subterranean nests (Forti et al. 2006). Cherrett and Peregrine (1976) mention a correspondence between the latitudinal richness gradient and the probability of some LCA species in becoming serious pests in the south of the continent. In our results, the number of potentially detrimental LCA species for productive systems remains constant (2-3 spp.) along entire studied latitudinal gradient (Figure 1), and it was apparently independent of the species richness. However, we observed a change in the composition of the more common and abundant species, the potentially detrimental species for each productive system. The replacement pattern may be explained by ecological, physiological or historical differences. On the one hand, differences in vegetation composition along the entire gradient may be shaping the distribution of these species (Farji-Brener and Ruggiero 1994) and/ or in turn interfering with the ecological dominance relationship.
Differences in the physiological tolerance (e.g. humidity and temperature) of the symbiont fungus could also be an important limiting factor for the distribution of certain LCA species, because their nesting behavior to control these conditions usually varies differentially interspecifically (Bollazzi et al. 2008). For example, some Acromyrmex species, restricted to more subtropical and humid regions in northern Argentina, made underground nests (e.g. Ac. subterraneus, Ac. balzani), and the species found more to the south tended to make dome nests (e.g. Ac. ambiguus, Ac. lundii) or had the ability to place the nest according to the microhabitat conditions (e.g. A. lundii). These ecological and physiological restrictions are surely had an important role in the historical dispersion processes of this species, but also the geological process of the South American Paran a basin dynamics could have played a certain role. In this context, the transformation of natural habitats into plantation areas could favor the establishment of these high-risk species rather than the more restricted ones, as has been shown in previous studies (Urbas et al. 2007;Wirth et al. 2007).

Species associated with productive systems
Most of the reports of damage caused by LCA in productive systems in northeastern Argentina refer to forest plantations (Table 1). This situation may be due to the extent of forest plantations in this region and also because of the low structural complexity of these monospecific plantations that favors the presence and abundance of some few detrimental LCA (Fowler 1983b). Depending on the density and size of their nests, the age of the plantation and the environmental conditions at the time of the damage, these harmful LCA can rapidly eliminate important sectors of young plantations. Unfortunately, there is scarce and undetailed information on LCA damage on forest cultivated areas in this region. A study on plantations from Pinus spp. Corrientes province (Argentina) revealed that ants of the genus Acromyrmex caused a loss of 21% of the seedlings during the first 65 days of life (Cantarelli et al. 2006;Cantarelli et al. 2008). Acromyrmex lobicornis has been reported causing economic damage in forest plantations in Corrientes and Misiones provinces (de Coll 2003), in vineyards in western Argentina (Dagatti 2016;L.A.C., pers. obs.), and as a potential pest ant in Patagonian forest plantations (P erez 2009).
Even so, LCA can also be a problem in other productive systems, like crops. Rosado et al. (2012) reported that Ac. ambiguus affects peach crops in wetlands in R ıo Grande do Sul, Brazil. Ferreira (1998) reported several LCA species (Ac. landolti, Ac. heyeri, At. capiguara, At. bisphaerica, At. laevigata and At. sexdens rubropilosa) attacking rice plants, but their damage represents only a reduction of 1% of the grain production. Nevertheless, it is clear that not only agriculture and forestry systems can be affected by LCA, but also, grasslands and pastures in Argentina, Brazil and Paraguay (Cherrett 1986). A nest of At. capiguara can cut 21 kg of grass/day, that would be able to feed 1.23 oxen/day/ ha in a cattle regime in Brazil (Amante 1967).

Current and potential species distribution
The ENM allows us to establish the environmentally suitable areas of seven economically important LCA species in productive systems of Argentina. It is not surprising that the species most commonly reported in forest plantations (Table 1) are the same that showed the higher suitability values within the northeast of Argentina.
We found some interesting distribution patterns for several LCA species. Acromyrmex lundii was the most distributed species in northeastern Argentina. Its success in the Mesopotamia and low basin of the Paran a River (e.g. Mesopotamia) could be due to its capacity to survive floods, because as they nest at the base of trees, they can quickly move their nests up the trees. The presence of Ac. ambiguus in Argentina is mainly limited to the low basin of Paran a-Uruguay rivers, despite this it is an important economic problem in Salicaceae plantations in some areas of the low delta of the Paran a River (N. L. Jim enez, unpublished data). Our models show that southern Brazil and Uruguay are regions with higher probability of distribution for this species along with the northeast of Argentina, especially in sand areas usually in costal environments (Bollazzi and Roces 2007 A. F. S R., pers. obs.).
Acromyrmex crassispinus, which is an important pest in forest plantations in Brazil (Nickele et al. 2009), was uncommon in the northeast of Argentina, and is scarce or even absent in other regions of Argentina. Our models predict its potential distribution in the Mesopotamian region and in a large part of Misiones province, suggesting that it could be a potentially detrimental species for argentine forest plantations. The model of Ac. lobicornis (similar to that of Ac. striatus), the most widely distributed LCA species in Argentina, clearly reflects its disjunct distribution, which consisted of populations along the Monte ecoregion, and scattered populations mostly located in arid rocky mountain habitats within the Atlantic forest ecoregion. Intermediate populations are found across remnants of the Espinal (the most arid and shrubby ecoregion after the Monte) in C ordoba, Santa Fe, Entre R ıos and Corrientes provinces.
The low basin of the Paran a River seems to be the southern distribution limit of the Atta genus, with At. vollenweideri as the southernmost species ($ 34 S), while At. sexdens was exclusively found in subtropical areas. Interestingly, both species partially overlap in southern Corrientes province. Our model was similar to the models previously published for At. vollenweideri (Sabattini et al. 2017). However, these models focused on Entre R ıos province alone, while ours provided the entire potential distribution for this species that includes also the provinces of Corrientes, Santa Fe, Chaco and Formosa. Populations of At. sexdens present in Argentina could also be part of a complex of species or at least some geographically differentiated lineages, surely At. sexdens rubropilosa, as it is noted in the phylogeny of Bacci et al. (2009). Therefore, this model can only be taken as an approximation.
In general, most non-detrimental LCA species usually have more restricted distribution ranges (Fowler et al. 1989) as could be observed in the province of Misiones where most species are uncommon, scant and have a very restricted distribution range. The distribution of these small-range species is probably related to the lower variation in the environmental conditions (temperature and humidity) in the most natural and conserved forest habitats in the Atlantic Forest ecoregion where they occur. They would have a narrower range of tolerance to temperature and humidity than the most distributed species, as observed in Ac. hispidus (N. L. Jim enez, unpublished data). Additionally, the change between the Paranaense ecoregion and the southern ecoregions of Fields and Weeds and Pampa, characterized by having more open vegetation and pastures (Burkart et al. 1999), is marked by the Espinal ecoregion belt that acts as a transition zone that seems to be where the replacement of LCA species apparently occurs. This phenomenon was more evident in the Atta species and in the Acromyrmex grass-cutting species (e.g. Ac. heyeri). Farji-Brener and Ruggiero (1994) explain that this phenomenon is apparently indicate by the replacement of woody vegetation by grass vegetation, where the subtropical rainforest in Misiones represents an important barrier to the expansion of the LCA grass-cutting species.
Although the LCA species that were modeled have wider spatial ranges and can be found occupying different regions, our niche model shows that the studied species are mainly influenced by the same climatic variables (Figure 3). The niche overlapping rates were closer and higher for the Acromyrmex species than for the Atta species. This is probably due to their phylogenetic closeness relationship and not to their particular behavior of nest-construction with mounds to protect the fungus garden. However, the capacity to maintain a suitable microclimate inside the mound during winter has also been fundamental to the spreading of LCA species such as A. lobicornis, into colder environments (Farji-Brener et al. 2003). In this sense, among the bioclimatic variables, temperature is fundamental for the establishment of the nests of these ants (Bollazzi and Roces 2010). Among the variables that best explain the models, two were related to the temperature: the temperature seasonality (Bio04), which is a measure of temperature change over the course of the year, and mean temperature of wettest quarter (Bio08), which provides the mean temperatures during the wettest three months of the year. The Bio04 is very similar to the intra-annual temperature variables found by Farji-Brener and Ruggiero (1994) to be the principal determinants of LCA richness patterns in southern South America.
We find that variables related to precipitation are also relevant to explain the models. One is the precipitation of the warmest quarter (Bio18), which would be relevant to the widespread of species like Ac. lundii or Ac. lobicornis. This is a variable that provides total precipitation during the warmest three months of the year and is a variable that can be affecting the species seasonal distributions. Acromyrmex lundii inhabits naturally irrigated areas and is prone to floods, while Ac. lobicornis inhabits more arid areas. The precipitation of the driest month (Bio14) was relevant for species like Ac. ambiguus and Ac. crassispinus that inhabit more subtropical regions and could be more affected by humidity in terms of their distribution ranges. On the other hand, although the internal temperature of the nest can be controlled by the structure of the nest (the presence of mound), the external temperature can affect the foraging pattern of the ants and therefore the amount of leaves that they can collect for the fungus growing (Nobua Behrmann et al. 2017). Likewise, the precipitation could be affecting the relative humidity of the environment, the flooding regimes and vegetation abundance and richness, which would affect the distribution of LCA as has been seen on a local scale (Sendoya et al. 2014), or even the capacity of foraging (Farji-Brener et al. 2018).
In summary, our study showed the richness and distribution patterns of LCA species in northeastern Argentina. Misiones province was the richest in LCA species, clearly showing a decrease in the number of species as latitude increases along the latitudinal gradient. However, the number of high-risk LCA species for forestations was similar throughout the entire latitudinal gradient, but with a change in the potential detrimental species among regions. Thus, Ac. ambiguus and Ac. lundii tend to be important pests in Buenos Aires and southern Entre R ıos; Ac. lundii and Ac. heyeri in northern Entre R ıos and Corrientes; and mostly At. sexdens in Misiones provinces. We find that ENM is a useful tool to assess the environment suitability areas where LCA pest species are likely shift their distribution within the northeast of Argentina, the most important forestry industry region of Argentina. Therefore, these models could be important tools for decision makers in qualifying the potential spread of destructive pests and to identify those species that do not represent any risk. and the Foundation for the Study of Invasive Species (FuEDEI).

Disclosure statement
No potential conflict of interest was reported by the authors.