Tree-dwelling ant survey (Hymenoptera, Formicidae) in Mitaraka, French Guiana

ABSTRACT Ants constitute a substantial part of the arthropod biomass in rainforests. Most studies have focused on ground-dwelling ants, which constitute almost half of the diversity of the ant assemblage. We report here the results of the first survey of tree-dwelling ants in French Guiana on a plateau and in a swamp palm forest (Euterpe oleracea Mart.) in the Mitaraka Mountains. We were interested in seeing the effect of topography and geographic distance on species richness and composition and to gather information on the species distribution on tree trunks. The fauna of Mitaraka was compared with one from a site 350 km distant (Petit Saut). In total 105 trees were sampled (30, 30, 45 in the plateau and the swamp forests of Mitaraka, and in Petit Saut plateau forest, respectively). Arboreal ants were attracted using tuna and honey baits spread along a rope reaching an upper branch, except for the palm swamp forest where the baits were only placed at 2 m high. A total of 34, 13 and 22 species were observed in these three respective sites. Six of these species are new records for French Guiana. In Mitaraka Camponotus femoratus (Fabricius, 1804) and Crematogaster levior Longino, 2003 co-occurred on trees (parabiotic association) and were among the most common species, along with Crematogaster tenuicula Forel, 1904 which was found on other trees (species exclusion). The Mitaraka Mountains appeared more species rich and had a species composition distinct from Petit Saut. Topography also influenced ant species composition. Almost half of the species collected by the baitline method were exclusively foraging in the canopy.


INTrodUcTIoN
Ants are arthropods of major ecological importance in tropical and temperate ecosystems. comprising a single family and with around 15 000 described species, their diversity is in the same order of magnitude as birds (Barrowclough et al. 2016;Janicki et al. 2016); as social organisms, they are ubiquitous in tropical terrestrial environments and can be sampled with standardized protocols (Agosti & Alonso 2000). These are the main reasons why they are a manageable arthropod group for biomonitoring (Underwood & Fisher 2006). This contrasts with other arthropod groups such as coleoptera which are much richer in tropical forests (Basset et al. 2012). Many more ant species remain to be discovered in unexplored geographical areas or habitats difficult to sample, such as the soil or the canopy (ryder Wilkie et al. 2007(ryder Wilkie et al. , 2010Ward 2010).
The Amazonian Basin remains one of these little explored regions. Most surveys of local richness have focused on grounddwelling ant communities. For example, in five localities across Guiana, Lapolla et al. (2007) found 230 species in the leaf-litter with 38-84 species per site. At the Nouragues research Station, French Guiana, from the base to the summit of inselbergs, Groc et al. (2009) found 196 species in the leaf litter and put in evidence that ant diversity was influenced by habitat type. In the Brazilian state of Acre, Miranda et al. (2012) sampled 222 ant species on the ground with pitfall traps and 115 ant species by beating the vegetation. Near Manaus, Brazil, Vasconcelos et al. (2003) also demonstrated the influence of topography, observing more species in river valleys than on forested plateaus. In the same area, at a larger scale (along a 2 000-km-long gradient), Vasconcelos et al. (2010) demonstrated the effect of rainfall and flooding regimes on species turnover. In another survey, using baits on the ground and on the vegetation, Vasconcelos & Vilhena (2006) showed a stratification in the ant assemblage. This stratification was even more thoroughly documented in a study by ryder Wilkie et al. (2010) in Amazonian Ecuador. These authors observed distinct ant assemblages in the soil, on the ground surface, and in the understorey and canopy. of the 489 species collected, 282 were found only on trees. These ants were obtained by fogging sampling by Terry Erwin who also worked in other parts of Amazonia. In Amazonian Brazil, his fogging campaigns revealed that ants were the most abundant arthropods in trees, representing over one third of the collected individuals and biomass (Erwin 1983;Adis et al. 1984). Wilson (1987) found in Erwin's samples in Peru up to 43 ant species from 26 genera from a single tree, and Tobin (1997) found 85 species from 29 genera from two trees and 11 associated vines for another site in Peru.
The aim of the current study was to complement these studies by: 1) surveying tree-dwelling ants in one of the most remote parts of the Amazon, the Mitaraka Mountains in French Guiana; 2) studying the effect of topography on arboreal ant assemblages; 3) studying the species turnover of tree-dwelling ant assemblages separated by large geographical distances; and 4) gathering information on vertical species distribution along trees.

Study SiteS Mitaraka
This species inventory is part of the "our Planet reviewed" ("La Planète revisitée") French Guiana 2014international biotic survey (Pascal et al. 2015Touroult et al. 2018). It took place in the Mitaraka mountain range bordering Suri-name and Brazil. Two contrasting forest sites were sampled: a plateau ("forêt de plateau"); and a swamp forest with with the palm tree Euterpe oleracea Mart. at the bottom of the valley ("pinotière"). Both were centred on a plot from the diadema ("dIssecting Amazonian diversity by Enhancing a Multiple taxa Approach") project (http://www.labex-ceba.fr/en/6337-2/). Their latitude and longitude were (2.233°N, -54.444°W) and (2.234°N, -54.448°W), respectively. The diadema project also includes ant samplings on the ground and in the understorey, relying on Winkler extraction, pitfall trapping and vegetation beating. These results have been reported elsewhere (Fichaux 2018). In the plateau forest, 30 tree canopies inside a circular plot 30 m in radius were sampled (area 0.28 ha, Fig. 1). In the swamp forest, they were collected from 30 trees within a 30 × 40 m plot (0.12 ha, Fig. 2). Ants were sampled between the 23 rd of February and 8 th of March 2017.

Petit Saut
To compare the faunal composition and diversity of Mitaraka, we used the data from an earlier survey in a plateau forest near the Petit Saut dam (Zone de relâcher, 5.068°N, -52.980°W) between the 13 th and 25 th of october 2010. A plot of 40 × 70 m (0.28 ha) was delineated there and 45 canopy trees   were sampled. This plot was centred around a large Azteca cf. chartifex colony to map the spatial extension of its territory. This site was c. 350 km away from Mitaraka but part of the same Amazonian forest block.

Ant SAmpling
We used the arboreal baitline protocol to collect tree-dwelling ants (Leponce & dejean 2011). The baitline protocol consists in putting a rope over the uppermost branch in the canopy using a sling shot (Sherrilltree® Big Shot). Baits are spread every 5 m along the rope, from 2 m above ground up to the uppermost branch, to detect species vertical stratification. Because the rope forms a loop, baits can be easily collected. Baits consist of a mixture of tuna (in vegetable oil) and honey. These baits represent a source of proteins, lipids and carbohydrates. The mixture is wrapped inside a paper towel and tied to the rope (see Fig. 3d, F). Ants dig inside the bait and remain on it even when it is brought back down for inspection. Baits were set in the morning and collected approximately four hours later, in the afternoon. In the event of heavy rain, baits were removed and replaced the next day. In the swamp forest, the Big Shot could not be used (no firm soil to anchor the pole of the sling shot) and baits were only put 2 m above the ground. A total of 99, 30 and 146 baits were set on the plateau, the swamp and Petit saut forests respectively. This method was designed to collect numerically dominant ants but it also collects part of the other species nesting on trees or at ground level. This method also has the advantage of allowing the vertical distribution of ant species along tree trunks to be assessed.

StAtiSticAl AnAlySeS
Species rarefaction curves were plotted on the species occurrences data matrices using EstimateS 9.1.0 software (colwell 2016) with 100 randomizations of the sampling order without replacement. Trees were considered as sampling units. The number of individuals collected on baits was not taken into account. The rationale is that ants are colonial insects and that it is only the number of colonies that is ecologically meaningful (Longino 2000; Leponce et al. 2004). If an ant species was found on a single or on multiple baits on the same tree its occurrence was counted as 1. To standardize the measurements of species richness between sites, an interpolation to a common number of occurrences (40) was carried out as well as the calculation of the chao2 estimator of species richness. The non-overlapping of 95% confidence intervals constructed from unconditional variance estimators was used as a conservative criterion of statistical difference (colwell 2016). For the analyses of species co-occurrences designed to reveal positive (i.e., parabiotic association) negative (i.e., exclusion) associations between species, the sampling unit considered was each bait. The independence of the presence of two species was measured with Fisher's exact-tests. These tests were only conducted on the most frequent species (i.e., present each on at least 10% of the baits).
The compositional similarity between two sites was measured with the chao-Jaccard index, which is an appropriate measure in the case of incompletely sampled fauna (chao et al. 2005).

Species richness
A total of 40 tree-dwelling species comprising 15 genera were found at Mitaraka (703 specimens collected). Five species belonging to the genus Pheidole Westwood, 1839 were new records for French Guiana (Table 1). The plateau forest was richer than the swamp forest: 34 vs 13 observed species (for 74 and 44 species occurrences, respectively). This difference was significant when considering a standardized species richness: 23.2 ± 6.2 vs 12.5 ± 5.5 species (rarefied richness for 40 occurrences ± 95% confidence interval) ( Table 1; Fig. 4). This is valid even when considering only baits collected near the ground at both sites because the accumulation of species in the plateau forest was similar for ground-and canopy-dwelling species (Fig. 4). The Mitaraka plateau forest was also richer than the distant forest used as comparison (Petit Saut). This might be due in part to the fact that the Petit Saut plot was largely occupied (19/45 trees) by a single colony of a numerically and aggressively dominant ant (Azteca cf. chartifex Forel, 1896). These Azteca Forel, 1878 were present in massive numbers on baits and excluded most other ant species from their territory. The chao2 estimator indicated that we possibly collected only half of the species present at Mitaraka with a total of 76.6 ± 29.2 and 36.7 ± 20.2 species expected in the plateau and swamp forest, respectively.
The faunal similarity between the two Mitaraka sites (chao-Jaccard: 0.372) was higher than with the Petit Saut site (0.136 or 0.047 for plateau and swamp comparisons, respectively) ( Table 1). Among the common species from Mitaraka, only Crematogaster tenuicula Forel, 1904 was found in abundance at Petit Saut as well.

Species vertical distribution
These data were only available for the plateau forest at Mitaraka. There, 15 out of 34 species were found up to the canopy (height ≥ 12 m) and most of them from the ground to the canopy (between 2 and 22 m, n = 10) ( Table 1). The other 19 species were found only above ground (between 2 and 7 m) but were species for which only one or two occurrences were noted. Species from Pheidole and Solenopsis were  Table 1. -List of species found at Mitaraka, in the plateau and swamp forests, and in the comparison site Petit Saut. Values represent the number of trees on which each species was observed (with the number of specimens examined under brackets). The minimum and maximum heights at which each species was collected along trees are also provided (in meters above ground). All the species are native to French Guiana, except the exotic Monomorium floricola (Jerdon, 1851), signalled with a *. Full details of examined specimens is provided in Appendix 1.

dIScUSSIoN
The plateau forest at Mitaraka appeared rich in tree-dwelling species compared to the swamp forest and Petit Saut. The species richness and composition was also different between Mitaraka and Petit Saut. These differences might be amplified in part due to the presence of a large colony of Azteca cf. chartifex in the Petit Saut plot, which decreased the diversity of ants found at baits due to exclusive competition known for territorially dominant arboreal ants (dejean et al. 2015). Altogether, 40 species were collected at Mitaraka but according to the chao2 estimator, the tree-dwelling ant fauna attracted to baits might actually be twice as rich. This seems plausible since in the Ecuadorian Amazon, different types of baits at ground level collected 83 of the 269 ground-dwelling species (31%) and 17% of the local ant fauna (489 species) (ryder Wilkie et al. 2010). during the same study 282 species were collected from the canopy by a very intense fogging programme spanning over height years. one might expect that at Mitaraka the use of various canopy sampling methods would increase the estimate of total number of tree-dwelling species. Insecticide fogging might be an option but it is time consuming in the field (i.e., it requires heavier equipment, and the weather conditions must be dry, warm and calm; Adis et al. (1998)) and in the laboratory (i.e., a huge number of individuals must be processed; e.g. 113 000 ants in ryder Wilkie et al. (2010)). Furthermore fogging does not provide information on vertical stratification as ants from different levels fall in the nets. other canopy sampling techniques (e.g. branch clipping, canopy pitfalls, traps, vegetation beating) usually require climbing which is difficult and risky (Yusah et al. 2012;Yusah et al. 2018). The newly developed arboreal baitline technique has the advantage of allowing researchers to collect from numerous trees quite rapidly, to detect the dominant ants in the canopy and to assess the distribution of species along the tree trunk.
The baiting method mainly attracted species with a high recruitment rate (i.e., Crematogaster, Pheidole, Solenopsis, Wasmannia). Wasmannia auropunctata (roger, 1863), a notoriously invasive species, is native to this zone (orivel et al. 2009). By contrast to other species of the genus Neoponera, almost all generalist predators, Neoponera commutata (roger, 1860) belongs to a group of species (with N. marginata (roger, 1861) and N. laevigata (Smith F., 1858)) which is specialized in raiding termite nests in the ground (Wheeler 1936;MacKay & MacKay 2010). other specialized ants were found, such as Cephalotes atratus (Linnaeus, 1758) and other species of the same genus which are known to eat anemophilous pollen that sticks the leaf surfaces (Baroni Urbani & de Andrade 1997). Almost half of the species collected by the baitline method were foraging in the canopy. The other half were found 2 m above ground and may correspond to groundnesting ant species occasionally foraging at the base of trees. Interestingly, some Pheidole and Solenopsis, two hyperdiverse genera very common at ground level, were found foraging high in the trees and may correspond to species nesting in suspended soil or epiphytes (Klimes 2017).
The new records for French Guiana were found in the genera Camponotus and Pheidole, which are among the main hyperdiverse genera of ants (Wilson 1976(Wilson , 2003. historically French Guiana has been very irregularly inventoried for its diversity of ants. Although there are many specimens of this fauna in the collections of Europe and North America, it remained poorly studied until the 1980s when a much larger research effort was applied to its rich biodiversity. The identification of species belonging to hyperdiverse genera remains a problem in tropical ant taxonomy. Many taxa (alpha taxonomy) were reliably described in the 19 th and early 20 th centuries (e.g. many medium to large species of the genus Camponotus), but the capability to identify many smaller species needs to be based on necessary extensive generic revisions in order to eliminate many suspected synonyms. This is in particular the case of Pheidole which was reviewed by Wilson (2003) although very little biological material from French Guiana was available in the collections he studied. This is the main reason why several Pheidole species found at Mitaraka are new records for French Guiana. At Mitaraka, Camponotus femoratus and Crematogaster levior were common and numerically dominant. These species are known to share the same nests (parabiosis), which are in antgardens, and this association seems to be favourable to both species (Vantaux et al. 2007;Menzel et al. 2014). Moreover, they also excluded the other most common species, Crematogaster tenuicula.
To conclude, our results suggest that the arboreal-dwelling ant fauna at Mitaraka is rich and with a composition possibly different from other parts of Amazonia. It was dominated by two parabiotic species: Camponotus femoratus and Crematogaster levior. of particular interest will be the comparison of this arboreal-dwelling assemblage with the ground-dwelling ant assemblage (sampled with the diadema protocol) to investigate if similar patterns of species turnover between sites are found.