Sampling a biodiversity hotspot : the orchid-bee fauna ( Hymenoptera : Apidae ) of Tarapoto , northeastern Peru , the richest and most diverse site of the Neotropics

The orchid-bee fauna of the region of Tarapoto, northeastern Peru, was surveyed using seventeen different scents as baits to attract orchid-bee males. Six hundred and fifty-nine males belonging to 41 species were actively collected with insect nets during 120 hours in late July and early August, 2012. Euglossa dressleri Moure, 1968, Euglossa laurensi Bembé, 2008, and Euglossa maculilabris Moure, 1968, three species belonging to the Euglossa cybelia species-group, are here reported for Peru for the first time. Previous sporadic and unpublished samplings in the area recorded eleven additional species. With 53 species, the region of Tarapoto can be considered the richest single site in the Neotropics for orchid bees. Diversity, estimated with the Shannon-Wiener diversity index (H’ = 3.02), was also the highest ever recorded for orchid bees.


Introduction
Orchid bees (Hymenoptera: Apidae: Euglossina) are important pollinators in Neotropical forests and considered to be key organisms in the ecosystems where they live (reviewed by Dressler (1982a)).These bees have become a favourite in ecological bee studies (e.g.Nemésio and Silveira, 2006b, 2007a, 2010;Rasmussen, 2009;Abrahamczyk et al., 2011) due to the ease of collecting their males, which are strongly and readily attracted to synthetic aromatic scents that mimic natural floral fragrances (Vogel, 1966;Dodson et al., 1969).
Although many orchid-bee samplings have been carried out in Peru and around 10% of all valid orchid-bee species have Peru as their type locality (see Nemésio and Rasmussen, 2011), there are only four published surveys of orchid-bee fauna in Peruvian territory, all carried out in lowland areas in the Amazon Basin (Pearson and Dressler, 1985;Bembé, 2002;Rasmussen, 2009;Abrahamczyk et al., 2011).
The region of Tarapoto, Department of San Martín, northeastern Peru (Figure 1), is of particular interest due to its singular location, exactly in the transition between lowland rainforest and lower montane rain forest, or cloud forest in Peru.The high precipitation associated with the abrupt elevation gradient in the region of the 'Cordillera Escalera' results in great environmental heterogeneity and a consequent unique biological diversity (Gallusser, 2002;Rasmussen, 2009).Data on the orchid-bee fauna of Tarapoto is available only from a single day of collecting, when 92 specimens belonging to 26 species were recorded (Abrahamczyk et al., 2011).This high richness revealed through a single-day survey, in addition to the fact that at least one rare and apparently geographically-restricted species has been described from the region (Rasmussen and Skov, 2006), suggested that the region of Tarapoto deserved further attention concerning its orchid-bee fauna.
The main goal of this study was to survey the orchid-bee fauna of the region of Tarapoto at an elevation gradient, focusing on the best preserved forest patches situated towards the higher parts of the 'Cordillera Escalera'.

Study sites
This study was conducted in forest patches at different elevations at the 'Cordillera Escalera', northeast of Tarapoto in the province and department of San Martín, northeastern Peru (Figure 1).Six sites were selected from 400 m to 1,000 m a.s.l.Samplings were carried out from July 20 th to August 1 st , 2012.

Data analysis
Diversity was estimated with the Shannon-Wiener diversity index (H'), as H' = -Σ 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 through the formula E = H'/ ln (S), where S is the species richness.Spearman rank correlation tests were used to estimate the effect of altitude on bee richness and diversity.The similarity in faunistic composition among all six sites was estimated by the similarity index of Morisita, recommended by Wolda (1981) for small samples.Similarity indexes that take into account not only presence of species, but also their relative abundances, are strongly recommended by Balmer (2002) since theoretically they better reflect natural processes.Based on those similarities, the areas were grouped using UPGMA (Sneath and Sokal, 1973).Only the collections carried out in 2012 were considered for the above analyses since they were performed under the same protocol.Similarity was also estimated for Peru as a whole (Morisita index and resulting similarities grouped using UPGMA).Data from the six sites sampled in this study in Tarapoto were merged and then compared to previous samplings carried out in Peru for which data on the relative abundance of each species was available.We used data from 13 sites: Tambopata Reserve, Madre de Dios, (Pearson and Dressler, 1985), Iquitos (Rasmussen, 2009) and Nanay, Tapiche river region (two sites), Moyobamba, Contamana, Pijuayal, Panguana, Cicra, Puerto Maldonado, and Tarapoto (Abrahamczyk et al., 2011).Since there were two samplings for Tarapoto (ours and that by Abrahamczyk and colleagues), we carried out three analyses: one including our data for Tarapoto, the second including Abrahamczyk and colleagues' (2011) data for Tarapoto, and the last one including both sets.

Results
Six hundred and fifty-nine orchid-bee males belonging to 41 species were collected in all six sites in 2012 (Table 1), whereas 28 species were collected at site-6 from 2002 to 2003 (Table 2).Data on abundance and, consequently, diversity and evenness were only available for the collections carried out in 2012.Richness, diversity and evenness were the lowest at the site situated at the lowest elevation and closer to the urban area, and were the highest at the sites situated at the highest elevation and best preserved vegetation (Table 1).No correlation was found between elevation and richness or diversity, even excluding site-1, the most distinctive of all and located outside forested areas.A positive correlation, however, was found between elevation and abundance (r s = 0.83, P < 0.05).
Eulaema nigrita Lepeletier, 1841 was the most common species in the region, but its dominance is strongly based on its abundance at site-1, where it represented more than 50% of the collected specimens.Euglossa orellana Roubik, 2004a was the most common species at sites 3 and 6, Euglossa ignita Smith, 1874 was the dominant species at site 2, Euglossa imperialis Cockerell, 1922 was the most common species at site 4, and three species, Eulaema marcii Nemésio, 2009, Euglossa amazonica Dressler, 1982d, andEl. nigrita were the most common species at site 5 (see Table 1).Three specimens of Eulaema tenuifasciata (Friese, 1925), two of El. meriana (Olivier, 1789), one of El. peruviana (Friese, 1903), and one of Eg. ignita were attracted to and collected on Anthurium flowers.
Fifteen out of the 17 scent baits were attractive to male orchid bees (Tables 3, 4).Only linalool and tricosane were not attractive.Eight scents (Table 3) attracted more than 10 specimens, and cineole, skatole, vanillin, methyl salicylate, methyl trans-cinnamate and eugenol were the six most powerful attractants in Tarapoto, responding for 90% of all collected specimens (Table 3).Methyl trans-cinnamate, skatole and methyl salicylate attracted the highest number of species (15), followed by cineole and vanillin (13 species each) (Table 3).Thirteen species were exclusively attracted to one specific scent, but eight of them were represented by singletons.Only five species represented by more than one specimen were exclusively attracted to one scent: all six Euglossa rugilabris Moure, 1967 3, 4).Only one species (Euglossa sp. 2) was only attracted to a lesser (weak attractive) scent (β-ionone, Table 4).
The ordination of the sites according to the similarity of their faunas showed medium to high overall (55% to 86%) values among the five forest areas and a relatively very low (34%) value between these areas and the open area situated at the lowest elevation (see Figure 2).
When the samplings from Peru were compared, the local fauna of Tarapoto recorded in the present study showed to be close to no particular area, sharing only ca.45% percent of similarity with all other areas, except the two sites close to the Tapiche river, Loreto, NE Peru, which shared about 35% similarity with all other Peruvian areas, Tarapoto included (Figure 3a).The situation is almost identical when data from Tarapoto was extracted from Abrahamczyk et al. (2011), the only difference being the value of similarity shared between Tarapoto and the remaining sites, which was slightly lower (Figure 3b).When both datasets from Tarapoto were included in the analysis, they revealed to be the most similar to each other, sharing about 42% of similarity with the remaining sites, except for the sites close to the Tapiche river (Figure 3c).

Faunistics, richness and diversity
The strategy of intensive sampling, using the same protocol as employed here, over a few days during the season when orchid bees are most actively foraging, has been demonstrated to be adequate for surveys focusing on this group of bees (Nemésio, 2010, 2011a, b, Nemésio et al., 2012).In fact, some authors argue that for orchid bees, surveys of a single day have great utility, and may reveal almost as much about local community structure as studies lasting a full year (Roubik, 2004a).Indeed, the grouping of both ours and Abrahamczyk and co-workers's (2011) data on the orchid-bee fauna of Tarapoto together among more than ten areas, even with different sampling protocols, seems to corroborate this point of view.
The species richness observed in Tarapoto is astonishing and only rivalled by that observed in central Panama, where 44 species were attracted to baits and six additional ones were recorded, totalling 50 species (Ackerman, 1983).The total of 53 species observed for Tarapoto region has no parallel in the Amazon Basin.In the western Amazon of Brazil, orchid-bee samplings usually record from 16 to 38 species (Powell and Powell, 1987;Becker et al., 1991;Morato 1994;Oliveira and Campos, 1996;Nemésio andMorato, 2004, 2006;Storck-Tonon et al., 2009, 2011) and even in the Peruvian Amazon, the highest richness ever recorded was that of the Tambopata Reserve, with 39  (Dressler, 1985;Pearson and Dressler, 1985).Other authors recorded 33 species for Huánuco (Bembé, 2002) and Iquitos (Rasmussen, 2009) regions, and Abrahamczyk et al.
(2011) recorded from 16 to 27 species in 10 different sites in eastern Peru.Also, the diversity recorded for Tarapoto, as estimated with the Shannon-Wiener diversity index (H' = 3.02), is the highest ever observed for orchid bees all over the Neotropics.
The high proportion of species represented by singletons in each of the six sites, and even when all six sites are considered together (Table 1), though, suggests that the orchid-bee fauna of Tarapoto may have been undersampled (see Coddington et al., 2009).In fact, previous samplings in the area by one of us (CR) and the one-day sampling carried out by Abrahamczyk et al. (2011) revealed eleven additional species not recorded in the present study.Ad hoc explanations for their absence can be argued for some, but not for all missing species.Species of Eufriesea Cockerell, 1908 are highly seasonal (see Kimsey, 1982) and since the collections in 2012 were carried out during a short period of time, the most probable explanation for the absence of Ef. ornata, Ef. purpurata, Ef. superba and Ef.surinamensis is that adults were not active in late July and early August, 2012.Exaerete dentata is usually not attracted to the scents ordinarily used as baits (see Nemésio and Silveira, 2006a), as well as El.polyzona.Although widely distributed throughout the Amazon, Eg. cognata and Eg.parvula are not common species in orchid-bee samplings, which might explain our failure in recording them in our 2012 survey.Euglossa laevicincta, Eg. modestior and El.cingulata were singletons in Abrahamczyk et al.'s (2011) sampling.These may be rare local species and were not detected in our samplings, although E. laevicincta was also reported from Tarapoto by Ramirez et al. (2010).Finally, we cannot explain the absence of Eg. crassipunctata during the collections carried out in 2012, since it is a species relatively common where it does occur.It should be emphasised that traditional orchid-bee inventories since the early 1970's are strongly based on attracting males to scent baits.It is known, however, that some species are usually not responsive to the scents commonly used in orchid-bee studies, especially some species of Euglossa Latreille, 1802 belonging to the subgenus Euglossella Moure, 1968(see Moure, 1996), almost all species of Eulaema Lepeletier, 1841 belonging to the Eulaema seabrai Moure, 1960 species group (see Nemésio and Silveira, 2004, 2006c, Nemésio and Ferrari, 2012), and the species of Exaerete Hoffmannsegg, 1817 belonging to the Exaerete dentata (Linnaeus, 1758) species group (see Nemésio and Silveira, 2006a).Eulaema tenuifasciata, for example, was only collected in the present study on Anthurium flowers.These constraints were recently discussed by Nemésio (2012) and alternative and complimentary sampling methods must be employed if one intends to better understand the actual richness of orchid bees in a given place.The alternative methods (e.g.collecting on flowers, searching for nests, use of trap nests, among others), however, are more time consuming, usually resulting in fewer specimens and are usually neglected when samplings are carried out during a limited period of time.In these circumstances, baiting, in spite of its limitations, offers the best results (see Nemésio, 2012).
Three species deserve special attention since they were recorded for the first time in Peru and, thus, represent range extensions of their previously known geographic distributions.Euglossa maculilabris Moure, 1968 was described from Panama and had additionally only been recorded in Costa Rica, Colombia and Ecuador (by Ramírez et al., 2002).The two specimens collected at the site situated at the highest elevation in Tarapoto represent a significant extension southwards from its southernmost previously known record in Ecuador.A single male of Euglossa dressleri Moure, 1968, also described from Panama, was recorded at the site situated at the lowest elevation in Tarapoto.This species had only been recorded in Costa Rica and Ecuador, besides the type locality (Ramírez et al., 2002).Euglossa laurensi Bembé, 2008 was recently described from Bolivia and this is the first record of this species outside Bolivia, extending its known distribution northwards.It should be pointed out that all three species mentioned above belong to the Euglossa cybelia species group, and a fourth species, Euglossa ioprosopa Dressler, 1982b, was also recorded in Tarapoto in four of the six sampled sites (Table 1).
Also, a single male of Euglossa retroviridis Dressler, 1982b, was recorded in the present study and it is the second specimen belonging to this species ever recorded in Peru -the other one recently reported by Abrahamczyk et al. (2011) from Nanay, Loreto.Finally, a single male highly similar to the recently described Euglossa williamsi Hinojosa-Díaz and Engel, 2011b was collected in the present study.This species was described based on only two specimens, one from Ecuador and the other one from central Peru.The specimen collected in Tarapoto is midway between both previously known geographic sites where this species is known to occur.Although this particular specimen has wider paraocular ivory markings than the other two known specimens (see Hinojosa-Díaz and Engel, 2011b), we consider it as just an intra-specific variation since it matched the original description in all other characters.Of course, larger series are necessary to understand the range of variations in this species and to be confident of the identity of our specimen.
The high richness here observed for orchid bees in Tarapoto is not entirely surprising, since the same phenomenon has already been recorded for other taxonomic groups (Davis, 1986;Dasmahapatra et al., 2010).Dasmahapatra et al. (2010) argued that the region of Tarapoto is a well-defined 'suture zone' between two areas of butterfly endemism in the lowland rainforest: the Río Mayo/upper Río Huallaga valley systems on one hand (Departments of San Martín and Huánuco), and the lower Río Huallaga and Río Ucayali regions on the other hand (Departments of San Martín, Loreto and Ucayali).These two regions are separated by the 'Cordillera Escalera', a low (mostly < 1500 m altitude) easternmost extension of the Andes.Suture zones of this kind (see Dasmahapatra et al., 2010 and references therein) are argued to form the meeting place for separate biotas recently expanded from refuges, here from the putative 'Huallaga' (Western) and 'Ucayali' (Eastern) Pleistocene refuges.This phenomenon could also explain, at least partially, the high richness of orchid-bee species observed in this study, as well as of other organisms in previous studies (Davis, 1986;Dasmahapatra et al., 2010).In fact, for forest taxa with a lesser dispersal ability and vagility than orchid bees, even high levels of local geographic differentiation and diversification were observed, as for butterflies of the tribe Ithomiini and the subtribe Heliconiina (Lepidoptera: Nymphalidae) (Gallusser, 2002;Dasmahapatra et al., 2010).

Scent preferences
The differential attractiveness of the scents used in the present study is similar to that observed in a previous study in Peru with multiple (>10) scents: a few of them are highly attractive and most scents attract a low number of species and specimens (see Pearson and Dressler, 1985).It should be stressed, however, that two scents that were poorly attractive in Pearson and Dressler's (1985) study (methyl trans-cinnamate, six specimens and three species; skatole, three specimens and two species) were highly attractive in the present study (see Table 3).One possible explanation is that both scents are usually commercialised as powder.It is possible that differences in the way these scents were handled and offered to the bees respond for the astonishing differences in both studies.We dissolved the powders in common alcohol (92.8°) until saturation and then, immersed a piece of string in the resulting solutions.These strings were then placed hanging from branches in the bait station, the alcohol rapidly evaporated and they became highly attractive (vanillin was also submitted to the same procedure).This same protocol has been used by one of us in Atlantic Forest areas (eastern Brazil) with similar success (Nemésio, unpub. data).Some researchers have avoided using skatole since it does stink and its bad smell usually impregnates all field and personal equipment.Although this is true, skatole has proved to be one of the best attractants for orchid-bee males, both in this study and in surveys carried out in the Brazilian Atlantic Forest (Nemésio, unpub. data).We therefore strongly encourage its use in orchid-bee inventories, since some species are strongly attracted by skatole, and there are even those exclusively attracted by skatole at some sites (see Table 3).
Less attractive scents should also not be discarded.For example, p-tolyl acetate was not a major attractant in the present study, but most specimens of Eulaema peruviana were attracted by this scent.It was also one of the few scents attractive to a rare orchid-bee in the Brazilian Atlantic Forest (Nemésio et al., 2012) and should be considered in future orchid-bee studies.The same was observed by Pearson and Dressler (1985), when anysil acetate was not a major attractant, but was useful for attracting some of the rarer species at Tambopata region not attracted by any other scent.Some closely related species apparently did not compete for the same scents in Tarapoto during this study.For example, four species of Euglossa (Glossura) showed different scent preferences: E. ignita, E. imperialis, E. occidentalis, and E. orellana (see Table 3).On the other hand, other species that are intimately related, as all Euglossa (Euglossa) belonging to the analis and cybelia species groups, showed strong preference for cineol.There is no study to date suggesting that phylogenetically closelyrelated species share preferences for the same or similar scents.This hypothesis should be further investigated.
Finally, it is outstanding the differences in scent preferences observed for the same species in the present study and that by Rasmussen (2009) in the region of Iquitos, the closest site sampled for orchid-bees which data on scent preferences are available.The area studied by Rasmussen (2009) is typical lowland Amazonian Forest, contrary to this study.Differences in natural availability of scents in both areas are here hypothesised to explain the observed differences, although it should be considered that Rasmussen (2009) only used five scents (all of them also used in the present study), and some of the most powerful attractants in the present study (e.g., skatole, methyl trans-cinnamate) were not used in Iquitos.
Based on the results here presented, we encourage researchers to use a wider array of scents, more than the usual five or six, in future orchid-bee studies to better sample those odd little-responsive species that usually escape our attention, especially in highly species-rich areas (see also Nemésio, 2012).

Similarity among Peruvian areas
Similarity among orchid-bee faunas in different Peruvian areas showed no obvious pattern.Some areas in the northern portion of Peru, such as Nanay and Iquitos, grouped with southern areas, such as Cicra and Puerto Maldonado, respectively, which otherwise share few similarities regarding their phytophysiognomies.It should be emphasised, though, that available data came from studies based on different sampling protocols, using different scent baits, during different periods of time.Nemésio and Silveira (2007b) had already pointed out this kind of limitation when comparing orchid-bee faunas throughout the entire Neotropical region.Thus, more long-term studies in Peru would be important to verify the validity of the short-term studies for assessing the orchid-bee communities of all sampled areas in Peru, particularly those rapidly sampled by Abrahamczyk et al. (2011).These studies may be of importance to provide a first overall view and highlight possible hotspots for orchid bees.Nevertheless, they may also be not intensive enough to reflect the actual diversity in a given area, as pointed out by Coddington et al. (2009), and we strongly suggest that Peruvian areas should be further investigated concerning their orchid-bee faunas, given their potential to reveal new and little known species, and help us to understand the geographic distributions of many species.

Figure 1 .
Figure 1.Map of Peru showing the location of Tarapoto.

Table 1 .
Diversity, evenness, species richness and number of specimens of each orchid-bee species collected at six sampling sites near Tarapoto, Department of San Martín, northeastern Peru, after 20 hours of sampling in each site.See text for location of each site.

Table 2 .
List of orchid-bee species collected in the region of Tarapoto fromApril, 2002 to April, 2003.

Table 3 .
Number of specimens of each orchid-bee species collected at six sampling sites near Tarapoto, Department of San species