The orchid-bee faunas ( Hymenoptera : Apidae ) of two Atlantic Forest remnants in southern Bahia , eastern Brazil

The orchid-bee faunas of the ‘Parque Nacional do Pau Brasil’ (8,500 ha) and ‘RPPN Estação Veracel’ (6,000 ha), two Atlantic Forest remnants in the southern state of Bahia, northeastern Brazil, were surveyed. Seventeen chemical compounds were used as scent baits to attract orchid-bee males. Seven hundred and twelve males belonging to 20 species were actively collected with insect nets during 80 hours in February and April, 2009. Euglossa marianae Nemésio, 2011, the most sensitive orchid-bee species of the Atlantic Forest, was recorded at both preserves, though in low abundance. ‘RPPN Estação Veracel’ is the smallest forest patch where Euglossa marianae has ever been recorded.

The 'Hileia Baiana', however, experienced strong deforestation in the last 40 years, when it is estimated that billions of trees were cut off (Fundação SOS Mata Atlântica, 1993), resulting in a highly fragmented landscape, with only a few larger forest patches left, usually not exceeding 25,000 ha (Ribeiro et al., 2009).Since orchid bees are forest insects and many of its species are highly dependent on well-preserved forest areas (Roubik and Hanson, 2004), species endemic in this region are more prone to population declines and even to extinction.Considering that new species have been recently described from the area (Moure, 1996;Nemésio, 2011bNemésio, , c, d, 2012)), further samplings in the region are still needed to a better knowledge of the actual orchid-bee richness at this part of the Atlantic Forest.
In this study one area not previously sampled in southern Bahia, the 'Parque Nacional do Pau Brasil', a 8,500-ha national park in the municipality of Porto Seguro, had its orchid-bee fauna surveyed for the first time.Also, the 'Reserva Particular do Patrimônio Natural (RPPN) Estação Veracel', a 6,000-ha private preserve located in the neighbourhood of 'Parque Nacional do Pau Brasil', previously sampled by Melo (2005), was also sampled again for comparison.

Study sites
This study was conducted at the 'Parque Nacional do Pau Brasil' (PNPB), a 8,500-ha preserve situated in the municipality of Porto Seguro, in the state of Bahia, northeastern Brazil , and the 'RPPN Estação Veracel' (REV), a 6,000-ha preserve situated in the municipality of Santa Cruz Cabrália, ca.40 km northeastern from PNPB (Figure 1), in February (REV) and April (PNPB), 2009, when orchid bees are most actively foraging in the region (e.g., Nemésio, 2011a, b).The vegetation in the region is essentially dense Atlantic Rain Forest [Central Lowland Forest according to vegetational physiognomic classification by Thomas and Carvalho (2003)].Most of both preserves consists of secondary forests in various successional stages, since pastures, Eucalyptus, other crops and mainly cocoa plantations surround the area and occurred in some parts of the preserves.Pristine forest is still found in a small area at REV.More areas of pristine forest can be found at PNPB, which forest is better developed than at REV. Precipitation in the area is ca.1,300 mm/year (Thomas and Carvalho, 2003).

Sampling
Twenty hours of active sampling with insect nets were performed in each of two selected sites in each preserve, totalling 80 hours, following the methodology proposed by Nemésio (2010aNemésio ( , b, 2011a, b), b): PNPB site-1 (16° 31' 11" S, 39° 17' 09" W, ca.90 m above sea level) was sampled on the 15 th (from 11:00 AM to 2:00 PM), 16 th (from 9:00 AM to 3:00 PM), and on the 18 th (from 9:00 AM to 2:00 PM) of April, 2009; PNPB site-2 (16° 28' 56" S, 39° 13' 24" W, ca.90 m above sea level) was sampled from the 19 th , 20 th and 21 st of April, 2009, from 8:00 AM to 3:00 PM (except on the 21 st of April, from 9:00 AM to 3:00 PM); REV site-1 (16° 20' 07" S, 39° 07' 37" W, ca.90 m above sea level) was sampled from the 4 th to the 6 th of February, 2009, from 9:00 AM to 4:00 PM (except on the 6 th of February, from 9:00 AM to 3:00 PM); REV site-2 (16° 22' 09" S, 39° 06' 43" W, ca.110 m above sea level) was sampled from the 7 th to the 9 th of February, 2009, from 9:00 AM to 4:00 PM (except on the 7 th of February, from 9:00 AM to 3:00 PM).PNPB site 1 and REV site 1 were situated in the interior of the forest, in pristine areas, whereas PNPB site 2 and REV site 2 were located close to the edge, to better sample the different physiognomies at the preserves.At each site, 17 scent baits were placed ca.2.0 m apart from each other at about 1.5 m above the ground.These baits were made of cotton wadding soaked with one of the following substances, known or believed to be attractive to orchid bees: benzyl acetate, benzyl alcohol, r-carvone, 1,8-cineole, p-cresol acetate, dimethoxybenzene, eugenol, b-ionone, methyl benzoate, methyl trans-cinnamate, heneicosane, methyl salicylate, skatole, tricosane, p-tolyl acetate, vanillin, and a mixture (1:1) of methyl trans-cinnamate and p-tolyl acetate.Baits with cineole, the most volatile compound, were recharged every hour.Bees arriving on the baits during the sampling period were collected with insect nets and killed with ethyl acetate.They were labelled as belonging to the project "Euglossina da Hileia Baiana" and were deposited at the Entomological Collection of the Universidade Federal de Minas Gerais (UFMG), where they were numbered from 19282-55494 to 19444-55928 (specimens collected at REV) and from 19604-56564 to 19730-56584 (specimens collected at PNPB).

Data analysis
Diversity was estimated with the Shannon-Wiener diversity index (H'), as H' = -∑ p i ln (p i ), where p i is the proportion of total number of species made up of the ith species (Pielou, 1975).Evenness (E) was estimated with the formula E = H'/ln (S), where S is the species richness.The similarity in faunistic composition among all sites was estimated by the percent similarity index of Renkonen, recommended by Wolda (1981) for small samples.Based on those similarities, the areas were grouped using UPGMA (Sneath and Sokal, 1973).

Results
Seven hundred and twelve orchid-bee males belonging to 20 species in three genera were collected in the present study, two hundred and seventy-two specimens (in 18 species) at PNPB and four hundred and forty individuals (in 17 species) at REV (Table 1).Euglossa mixta Friese, 1899 (23% of all collected bees) and Eg.carolina Nemésio, 2009 (20%) were the most common species at PNPB and Eg.imperialis Cockerell, 1922 (22%) and Eg.carolina (17.5%) were the most common species at REV (Table 1).The only species represented by a singleton was Eg. pleosticta Dressler, 1982.The same sampling effort was carried out in all four sites, allowing a direct comparison.Abundance was higher at REV (ca.eleven specimens/hour) than at PNPB (ca.seven specimens/hour).Diversity and evenness were high and very similar among all sites.(H' = 2.20 to 2.25; E = 0.79 to 0.81) (Table 1).The ordination of the sites according to their faunas (Figure 2) showed a relatively low overall similarity among sites, but each site showed roughly the same similarity when compared to each other (ranging from 54% to 67%).Both sites at REV grouped together, with ca.67% similarity, and both sites at PNPB also grouped together, with some 63% similarity, and then both sets grouped with ca.60% similarity (Figure 2).REV-2 and PNPB-2, both sites situated at the forest edge, showed ca.64% similarity.

Ecology and faunistics
The efficiency of the sampling methodology used here was previously discussed in depth (Nemésio, 2010b(Nemésio, , 2011a, b;, b;Nemésio et al., 2012).The relatively low abundance found in the present study (seven to eleven specimens/ hour), compared to other studies carried out at the same region (Nemésio, 2011a(Nemésio, , b, 2013)), should be understood as natural of the sampled areas.Another large preserve, the 'Reserva Biológica de Una', also in southern Bahia, was sampled just a few days before the samplings at the 'RPPN Estação Veracel', with a higher abundance (14 specimens/ hour) (Nemésio, 2013).Similarly, samplings at the nearby 'Parque Nacional do Monte Pascoal' and adjacent forest patches some days before and after samplings at 'PN Pau Brasil' revealed a higher abundance (unpub.data).
Diversity, evenness and richness were very similar among all sampled sites.The overall similarity among sites (Figure 2), however, was not particularly high.In other studies in the Atlantic Forest, sites situated in different forest patches, even with urban matrix among them (Nemésio andSilveira, 2007a, 2010), presented similarities higher than those presented here.Also, sites sampled at the same forest patch in other studies (e.g., Tonhasca Junior et al., 2002;Nemésio and Silveira, 2006) showed higher similarities.On the other hand, the overall similarities reported here are similar to those found among sites at the nearby 'Reserva Biológica de Una' (Nemésio, 2013).This suggests that either the sampling effort, in spite of resulting in a high number of collected bees, is still not effective enough to thoroughly sample each site or that within-habitat heterogeneity (see Armbruster, 1993) should be considered when sampling areas with higher diversity -although Melo (2005), when sampling many sites at REV (see discussion below) found results very similar to those presented here.
A previous study had been carried out at 'RPPN Estação Veracel' (Melo, 2005), but for that unpublished Master's Thesis work, a different methodology was used: bait traps were used to attract and collect the bees, instead of active collecting with hand nets (present study) and, thus, comparisons should be made carefully.Nemésio and Morato (2006) have shown that bait traps introduce a bias, since the larger bees of the genus Eulaema tend to proportionally increase their numbers in such samplings, because the smaller Euglossa escape more easily.Mattozo et al. (2011) later confirmed Nemésio and Morato's (2006) results.Melo (2005) collected the same four species of Eulaema recorded in the present study (for differences in species nomenclature used in both studies, see Nemésio, 2009Nemésio, , 2011a) ) and Eulaema species represented 13.5% of the orchid-bee community, whereas they represented 9.7% in the present study (only REV sites considered, for comparison).
Table 1.Diversity, evenness, species richness and number of specimens of each orchid-bee species collected at sites 1 (forest) and 2 (edge) at both 'Parque Nacional do Pau Brasil' (PNPB) and 'RPPN Estação Veracel' (REV) in southern part of the state of Bahia, northeastern Brazil, after 20 hours of sampling at each site.Four of the five most common species in Melo's (2005) study were also the most common in the present study: Eg. imperialis represented 22% of the specimens in Melo's study (22.3% in the present study); Eg. ignita represented 21.8% in 2005 (15% in the present study); Eg. carolina, then considered as Eg.cordata (Linnaeus, 1758(Linnaeus, ), represented 17.1% in 2005 (17. (17.5% in the present study), and Eg.mixta Friese, 1899 declined from 11.4% to 5% in the present study.Melo (2005) collected 22 species at REV (17 species were collected in the present study), the difference was due to two species of Eufriesea Cockerell, 1908 (no Eufriesea species was collected in the present study) and four species of Euglossa collected in low abundances by Melo (2005) -counterbalanced by the collection of Eg. marianae in the present study, a species not collected by Melo (2005).

Species
The presence of Eg. marianae at REV is remarkable, because it is the smallest forest patch where this species was ever found, together with 'Estação Ecológica (ESEC) de Murici', in Alagoas (Nemésio, 2010b).As happened at ESEC Murici, the population of this species seems to be very tiny at REV and the only two recorded males were collected at the site in the interior of the forest, as usually preferred by this species (see Tonhasca Junior et al., 2002;Nemésio and Silveira, 2006).
The nine specimens of Eg. cognata Moure, 1970 collected at PNPB represent the northernmost record of this species in the Atlantic Forest.Previous studies at the 'Parque Estadual da Serra do Conduru' (Nemésio, 2011a) and at the 'Reserva Biológica de Una' (Nemésio, 2013), two areas some 100 km north of the areas sampled in the present study, failed to sample this species.Euglossa cognata seems to be a species dependent on well-preserved forest areas (see Nemésio and Silveira, 2006) and its absence from the neighbor REV [not collected by Melo (2005), either] supports this interpretation.
After comparing Melo's (2005) study and the sampling at 'Parque Estadual da Serra do Conduru', Nemésio (2011b) concluded that twenty-six orchid-bee species occurred in the region.Three additional species were recently recorded at 'REBIO Una' (Nemésio, 2013), raising the number of known species in this area to 29.

Final remarks
Similar to many other "protected" natural areas in northeastern Brazil, the situation faced by both preserves is problematic (few personnel to patrol the area; presence of hunters and wood dealers; legal problems with former owners of the land; invasion by "alleged" Indians who claim the area).As a private preserve belonging to a large company dealing with cellulose, REV maintains a professional team of park guards who patrol the area.Nevertheless, this area is more close to urban areas than PNPB, which makes it easier the access of hunters and wood dealers.On the other hand, PNPB is a public national park, which is virtually abandoned by the government, as happens to almost all Brazilian forested preserves belonging to the Brazilian government.Only four men were responsible for patrolling this 8,500-ha national park when I visited the area, an obviously insufficient number to adequately patrol such a large area.The presence of apparently sensitive orchid-bee species, such as Eg.marianae (see Tonhasca Junior et al., 2002;Nemésio and Silveira, 2006;Nemésio, 2010bNemésio, , 2011b)), Eg. roubiki and the recently described Euglossa pepei Nemésio and Engel, 2012, suggests that these areas, especially PNPB, still hold an important habitat for most, if not all, local orchid bees.

Figure 1 .
Figure 1.Map showing the exact locations of 'Parque Nacional do Pau Brasil' and 'RPPN Estação Veracel' in southern Bahia, eastern Brazil, and the exact locations of the sampling sites (black circles) within each area.