Are orchid bees at risk ? First comparative survey suggests declining populations of forest-dependent species

The two largest Atlantic Forest remnants in the state of Espírito Santo, eastern Brazil, namely ‘Reserva Biológica de Sooretama’ (REBIO Sooretama) and ‘Reserva Natural Vale’ (RNV), were surveyed for their orchid-bee faunas. Seventeen scent baits were used to attract orchid-bee males. Three-thousand, two hundred and twenty-five males belonging to 24 species were actively collected with insect nets during 100 hours in March, April and December, 2009. In comparison with a previous study in the same area twelve years before, it is evident that the abundance of all forest-dependent orchid bees analysed declined around 50%, and it was statistically significant (P = 0.022) for Euglossa marianae Nemésio, 2011, the most sensitive to anthropogenic disturbances of all Atlantic Forest orchid bees. On the other hand, the abundance of populations of species tolerant to open or disturbed areas rose. Possible explanations are discussed.


Introduction
Studies on the effects of habitat fragmentation on populations of orchid bees (Hymenoptera: Apidae: Euglossina) in the Amazon have shown that some species are much less tolerant to anthropogenic disturbances than others (e.g., Powell and Powell, 1987;Morato et al., 1992;Morato, 1994).If the same is true for orchid-bee species in the Atlantic Forest domain, a highly fragmented biome, the obvious implications would be that a suitable environment could become very limited for some species.Indeed, recent studies (Tonhasca Junior et al., 2002a;Nemésio and Silveira, 2006a;Giangarelli et al., 2009) have suggested that some Atlantic Forest species may be intolerant to disturbed areas and, thus, restricted to the few larger areas that still hold suitable habitats.Nemésio (2009) carried out the first attempt to evaluate the conservation status of each orchid-bee species occurring in the Atlantic Forest under IUCN criteria, and suggested that a few species should be included in some of the risk categories.More recently, an apparent local extinction was reported (Nemésio, 2011b) and additional surveys extended our knowledge on the actual geographic distributions of many of the rarest species (Nemésio 2010b, 2011a, e, Nemésio et al., 2012).
Nevertheless, it is not common to find data in the literature involving surveys in the same area at different times.Thus, it is impossible to evaluate changes in community composition over the years and assess if any species population are declining or not, particularly of those more sensitive species.Change in population size is one important criterion to estimate the possible fate of a species (locally or globally) according to IUCN risk categories (IUCN, 2001).
The state of Espírito Santo is an ideal and obvious candidate to try a "second survey" at the same site.Most of the original forest cover in Espírito Santo was wiped out and deforestation was particularly severe in the northernmost portion of the state, where only scattered fragments remain (Fundação SOS Mata Atlântica, 1993;Ribeiro et al., 2009), the largest of them, 'Reserva Biológica de Sooretama', not exceeding 24,000 ha (Ribeiro et al., 2009).This region, however, together with the southernmost portion of the state of Bahia, holds one of the most exuberant portions of the Atlantic Forest.More importantly, the oldest full survey in a large forest remnant was carried out at 'Reserva Natural Vale ' (22,000 ha) in 1996and 1997(Bonilla-Gómez, 1999).Together with the adjacent 'Reserva Biológica de Sooretama', these two areas represent around 10% of all remaining vegetation cover for Espírito Santo (Fundação SOS Mata Atlântica, 1993).
The main goals of this study were, thus: (i) re-sample 'Reserva Natural Vale' in order to estimate possible changes in its orchid-bee community, especially populations of the most sensitive species; (ii) sample for the first time the neighbor 'Reserva Biológica de Sooretama', in order to expand distributional and taxonomic knowledge on the orchid bees of the Brazilian Atlantic Forest, with special emphasis to the region of northern Espírito Santo, by providing relevant data on the orchid-bee species found in a poorly known area of one of the most threatened biomes in the world.

Study sites
This study was conducted at 'Reserva Biológica de Sooretama' (REBIO Sooretama), a 24,000-ha preserve belonging to the Brazilian government, situated in the municipality of Sooretama, and at 'Reserva Natural Vale' (RNV), a 22,000-ha particular preserve situated in the municipality of Linhares, in the state of Espírito Santo, southeastern Brazil (Figure 1), in March, April and December, 2009, during the summer in Brazil, when orchid bees are most actively foraging (e.g., Martins and Souza, 2005).Both preserves are connected through a narrow patch (in the easternmost end of 'REBIO Sooretama' and the northwesternmost section of RNV -see Figure 1).Thus, both preserves may be treated as a whole continuous 46,000-ha area, since many animals can use both areas freely (a federal road lies between both preserves, which limits the faunal interchange).Vegetation in this area is essentially dense Atlantic Rain Forest according to the vegetation map of Brazil (IBGE, 1993), or as lowland rain forest (Oliveira-Filho and Fontes, 2000), but 'REBIO Sooretama' is more compact, with a lower perimeter than RNV, which is roughly divided into three large blocks of vegetation with a very large perimeter (i.e., it suffers more pronounced effects around the border than 'REBIO Sooretama') (see Figure 1).'Reserva Natural Vale' is also closer to the coast than 'REBIO Sooretama' and there are more patches of vegetation situated on sandy soil, the socalled 'mussunungas', which consists in sparser and lower vegetation.The climate is wet tropical (Aw) with a rainy season in summer and a dry season in winter.Precipitation in the area is ca.1,200 mm/year, with a strong variability from year to year (Srbek-Araújo et al., 2009).

Data analysis
Diversity was estimated through Shannon-Wiener diversity index (H'), as H' = -Σ p i ln (p i ), where p i is the proportion of the total number of species made up of the ith species (Pielou, 1975).Evenness (E) was estimated through 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).

Comparison to previous study
Bonilla-Gómez (1999) sampled RNV for one year (September, 1996to August, 1997) with monthly samplings at each site from 08:00h to 13:00h.She selected three areas at RNV and three sampling sites in each area.Thus, nine sites, five hours for each site, monthly, during a whole year were sampled.It totals to 60 hours per site during one year, or 540 hours of sampling during her study.Bonilla-Gómez (1999) sampled (she did not collect all specimens) 16,177 specimens, or 30 specimens/hour.In the present study, one to two sampling sites in each one of Bonilla-Gómez's (1999) areas were selected.RNV-1 and RNV-4 corresponds to her 'Peroba Osso' site; RNV-2 corresponds to her 'Paraju' site, and RNV-3 corresponds to her 'Farinha Seca' site.Bonilla-Gómez (1999) observed two peaks of abundance and richness at RNV, one at the end of the rainy season (March-April) and the second one at the beginning of the rainy season (November-December), which are the exact seasons chosen in this work to sample the area, to make both data sets comparable.Bonilla-Gómez (1999) used five scent baits (cineole, eugenol, methyl-salycilate, skatole and vanillin), all of them are used in the present study.However, twelve more scents were used in this study, see above.In order to compare possible changes in the community composition, all species with well-defined habitat preferences were selected (those tolerant to open / disturbed habitats and those dependent on forest interior) based on previous studies (e.g.Morato, 1998 ;Tonhasca Junior et al., 2002a;Nemésio and Silveira, 2006a;Nemésio 2010bNemésio , 2011a, b), b).Euglossa mixta Friese, 1899 and Euglossa retroviridis Dressler, 1982 specimens in Bonilla-Gómez's (1999) study were grouped together as Euglossa mixta, since this distinction was previously considered as a misidentification (see Nemésio, 2009).The bees listed as Euglossa chalybeata Friese, 1925, Euglossa cordata (Linnaeus, 1758) by Bonilla-Gómez (1999) are here treated as Eg.marianae Nemésio, 2011b, Eg. roubiki Nemésio, 2009and Eg. carolina Nemésio, 2009, respectively.Two data sets were used: (i) the relative abundance of each species (when compared to the overall community), and (ii) the absolute abundance of each species (number of specimens/sampling effort in hours) in each sampling site.The hypothesis that the absolute abundance of these species changed through time was evaluated using t-tests.Data were square-root transformed prior to the analyses in order to meet the requirements of data normality.
In comparison to the previous study by Bonilla-Gómez (1999), it was revealed in this study that the abundance of orchid bee males (32.6 bees/hour, considering only RNV sites) was extremely similar to those previously found (30 bees/hour) at the same sites twelve years before -it raises to 38.9 bees/hour if only samplings in April and December are considered (Bonilla-Gómez, 1999).Eight species were considered to have their habitat preferences well-defined, four of them typically associated to open or disturbed habitats (or edges of well-preserved forest patches), and four species were associated to the interior of forests in well-preserved forest patches (see Table 2).All four species considered as forest dependent showed a clear tendency for a decline in their populations, both when considering their relative abundances (towards the overall orchid-bee community), and when considering their absolute abundance (number of specimens/sampling effort in hours).In contrast, three of the four species more tolerant to open or disturbed habitats showed a raise in their populations in both analyses (Tables 2 and 3).The only exception was Eulaema nigrita Lepeletier, 1841, which was considered to be a species tolerant to open habitats but which population declined (Tables 2 and 3).These abundance changes over time, however, were not statistically significant for most species (see Table 3).Nevertheless, abundance changes of Euglossa marianae, the most forest-dependent of all orchid bees in the Atlantic Forest, showed a significant value (P=0.022); an almost significant value (P=0.054) was found for Euglossa cognata, a close ally of E. marianae and also strongly associated to large, well-preserved forest patches.

Discussion
The efficiency of the methodology employed here, already used before with remarkable results (Nemésio 2010b;2011a, b;Nemésio et al., 2012), revealed that this strategy can equal the collection rates of traditional inventories with monthly samplings.Nevertheless, as most collections made in RNV (sites 1 to 3) occurred at the end of the wet season, when species of the highly seasonal genus Eufriesea were not active, the relative abundance of species of this genus is certainly underestimated.Bonilla-Gómez (1999) recorded five species of Eufriesea at RNV, whereas only two species were recorded in the present study.Bonilla-Gómez (1999) recorded twenty species of Euglossa, four of Eulaema (all recorded in the present study) and two of Exaerete (both recorded in the present study), totalling 31 orchid-bee species.Four species of Euglossa recorded by Bonilla-Gómez (1999) were not collected in the present study, but they may represent misidentifications, as discussed by Nemésio (2009).Furthermore, these additional species were represented by few (three or less) specimens.The bee treated as Euglossa magnipes Dressler, 1982 by Bonilla-Gómez (1999) is probably Eg. monnei Nemésio, 2012.Differences in nomenclature between Bonilla-Gómez (1999) and the  present study were already discussed by Nemésio (2009Nemésio ( , 2011aNemésio ( , b, 2012) ) and Nemésio and Rasmussen (2011), with the exception of the above mentioned Eg. monnei and the also recently described Eg.marianae (treated as Eg.analis in Bonilla-Gómez's study).
The dominance of Eg. carolina in all sampling sites is striking and deserves further consideration as this species has been associated to highly disturbed environments (discussed by Tonhasca Junior et al., 2002a;Nemésio and Silveira, 2006a;Nemésio, 2009Nemésio, , 2011a)), and it is one of the most common orchid-bee species in urban areas in Espírito Santo and Bahia (pers.obs.).Since 'REBIO Sooretama' and RNV are the largest and best preserved patches of forest in the state of Espírito Santo, it would be expected to find a more diverse orchid-bee fauna, with a predominance of forest-dependent species.Nevertheless, the overall evenness found at RNV is very low, particularly at RNV-2 and RNV-3, the sites located closest to the coast and on sandy soils.The results found here are similar to those by Bonilla-Gómez (1999), when Eg. carolina -treated as Eg.cordata in that study -represented almost 45% of the total community.A possible explanation for this apparent contradiction is that all of the surrounding areas of both preserves are made up of pastures and crop plantations, which are suitable environments for Eg.carolina.Moreover, the extensive perimeter of RNV, with its obvious edge effects, make this preserve particularly suitable for species typical of open habitats, since there are few areas of environment deep inside the forest.It is not surprising, thus, that Eulaema nigrita, another species typical of open habitats (see Morato, 1998;Tonhasca Junior et al., 2002a;Nemésio and Silveira, 2006a), was the second most abundant species in the present study -and the third most abundant in Bonilla-Gómez's (1999) study.A lower abundance of Eg. carolina and, particularly, of El. nigrita (only the sixth most abundant) in the neighbouring 'REBIO Sooretama' (see Table 1), a much more compact (which reduces edge effects) forest patch (see Figure 1), seems to corroborate the hypothesis above.Finally, another apparent paradox is also supportive of the hypothesis above: Eg. marianae, considered as a species highly sensitive and dependent on environments typical of deep forest (see Tonhasca Junior et al., 2002a;Nemésio and Silveira, 2006a;Nemésio, 2010bNemésio, , 2011a -in all these works treated as Eg.analis -and Nemésio 2011b), was the third most common species in the region -the second one in Bonilla-Gómez's (1999) study, and its relative abundance was higher in 'REBIO Sooretama' and the best preserved sites at RNV (sites 1 and 4).This apparent contradiction (the most tolerant and sensitive orchid-bee species are the most common ones in the same area) may suggest that these areas are the only ones large enough to hold viable populations of Eg. marianae, but being surrounded by open habitats, may have been "invaded" by a species highly tolerant to disturbed areas.It is very important to notice that the present study was carried out twelve years after the end of Bonilla-Gómez's work field, in 1997.Euglossa marianae represented, then, 17% of the total community of orchid bees (21% if only samplings in April and December are considered) and only 6.6% of the total community in the present study (see Table 1).Even in the sites where this species was most abundant ('REBIO Sooretama' and RNV-4), it barely reached 10%.If the data obtained in the present study is to be trusted, one must consider the possibility that this species declined roughly 50% of its contribution to the total community of orchid bees in 12 years, exactly what Nemésio (2011b) suggested to have happened in the smaller forest patches situated in the northern section of Espírito Santo, where this species seems to have declined to extinction in about 40 years.If both species compete for any kind of resources, it seems that Eg. carolina is a serious competitor and may put pressure on the survival of the Eg.marianae population at both preserves, particularly at RNV. Tonhasca Junior et al. (2002a) were the first to notice that Eg. marianae populations could be "trapped" in the largest forest patches in the Atlantic Forest, with no migration (Tonhasca Junior et al. 2003).Nemésio and Silveira (2006a) showed that this species has a strong preference for the inner areas of large forest patches, but it was only recently that it was warned that this species may be facing a real risk of extinction, since it has not been found in forest preserves smaller than 6,000 ha (see Nemésio, 2011b).In fact, Nemésio (2011b) was only able to confirm its presence in 15 areas in the Atlantic Forest, and suggested that it may be locally extinct in two of them,  (Roubik, 2001), it seems unreasonable to consider the situation of the Eg.marianae population as a simple matter of fluctuation over time.A more detailed comparison between Bonilla-Gómez's (1999) and the present study, focusing on species typically tolerant to open or disturbed habitats and on species typically dependent on environments in the interior of forests shows that all species typical of deep forest interiors declined their abundances, whereas most of those typical of open or disturbed habitats increased their abundances (see Tables 2 and 3).Thus, it doesn't seem to have been a phenomenon affecting only a particular species, but a series of species more sensitive to deforestation and consequent loss of suitable habitats.
It is important to stress, however, that the decline reported here was observed in a short 12-year time span.It may take some more time to be definitely noticed for other species, theoretically less sensitive than Eg.marianae.
More studies are obviously necessary to further investigate the hypothesis presented here.Only two studies, however, were carried out in large forest remnants (> 20,000 ha) more than ten years ago in the Atlantic Forest: that by Tonhasca Junior et al. (2002a), when samplings were carried out from 1997 to 1999, and the one by Nemésio and Silveira (2006a), when samplings took place from 1999 to 2000.Repetition of surveys in these areas, preferably at the same sampling sites, would allow a better understanding of the effects of habitat fragmentation on populations of highly sensitive orchid-bee species.

Table 2 .
Comparing relative abundance (to overall community) of species tolerant to open habitats and species dependent on environments in the interior of forests from two studies carried out at 'Reserva Natural Vale', in the state of Espírito Santo, eastern Brazil, with a time span of 12 years between both samplings.

Table 3 .
Comparing the absolute abundance of species tolerant to open habitats and species dependent on environments in the interior of forests from two studies carried out at 'Reserva Natural Vale', in the state of Espírito Santo, eastern Brazil, with a time span of 12 years between both samplings.Values represent the mean number (±S.E.) of bees captured per hour in each sampling site (n = 9 and n = 4, respectively, for the Bonilla-Gómez's and the present study).P-values are from t-tests (data were square-root transformed prior to the analyses).of this species could not be recently confirmed in two areas of northern Espírito Santo.Although populations of most insects, orchid bees included, tend to fluctuate over time