1 Introduction

In the face of extreme loss and fragmentation of habitats, the functional connectivity emerges as a critical variable for biodiversity conservation (Sanderson et al. 2003; Wiens 2007). The functional connectivity is the response of an organism to the structure of a landscape; thus, its evaluation requires information about the spatial configuration of the habitats and non-habitats or matrices, as well as about the pattern of movements of the organisms (Tischendorf and Fahrig 2000; Crooks and Sanjayan 2007).

The effects of a matrix tend to be species-specific, and a more heterogeneous matrix has more probability of facilitating movements of a wider variety of organisms (Kattan and Murcia 2003). Managing the matrix may be a more effective way of managing the landscape to preserve or restore the functional connectivity; however, for this to be possible, it is necessary to understand how species interact with matrices (Fahrig 2001, 2002; Wiens 2007).

Even the forest dwellers with high mobility may be affected by the highly fragmented stage of most rainforests, as is the case of the Brazilian Atlantic Forest (e.g., Guedes et al. 2005). Euglossina bees may probably respond more slowly to tropical forest fragmentation due to their high mobility. The biggest Euglossine bees can fly long distances and they can cross open areas including lakes or small bays in the sea (Janzen 1971; Roubik and Hanson 2004; Wikelski et al. 2010). However, high mobility per se may not grant functional connectivity depending on the species fidelity to the forest habitats and depending on the type and extension of anthropic matrices. Some anthropic matrices can reduce connectivity while others may offer resources or even be secondary habitats for some bees (e.g., Steffan-Dewenter 2003; Jauker et al. 2009; Ramalho et al. 2013).

In this study, we evaluate the permeability of oil palm, piassava palm, and rubber tree crops to the forest dwellers Euglossina bees of Brazilian Atlantic Rain Forest (BARF). These are the most extensive perennial crops and key anthropogenic matrices in the northern portion of the Brazilian Atlantic Forest Corridor (AFC). Species-specific responses are expected to each matrix. If the degree of forest fragmentation in the north of the AFC is already affecting the mobility of some of these bee species, as we expect to happen, we should also find variable responses to the different crop matrices.

2 Material and methods

2.1 Study area and sample design

The research was carried out in the Brazilian AFC, in the Camumu-Cabruca-Conduru (CCC) focal area. The AFC is a Brazilian initiative to preserve the Atlantic Rain Forest, one of the hottest “hot spot” of biodiversity in the world (Myers et al. 2000). The CCC is in the northern portion of AFC and comprises the mini-corridors Restinga (Sandbanks) and Papuã-Pratigi (Batista 2008; Online Resource 1) at Bahia State (Brazil).

Three landscape mosaics were chosen in the CCC, with native forest remnants and the three main regional crop matrices, in terms of extension of cultivated land area: (1) mosaic with the oil palm matrix (Elaeis guineensis Jacq.); (2) mosaic with the piassava palm matrix (Attalea funifera Mart.); (3) mosaic with the rubber tree matrix (Hevea brasiliensis Muell. Arg.).

The oil palm and piassava palm reach between 15 and 20 m high. They form perennial crops which demand little mechanization and reduced employment of agricultural pesticides (Guimarães and Silva 2012; Brazilio et al. 2012). The oil palm tree is exotic and, in general, it is planted with 7.8 m spacing between-lines and 9 m between-plants in lines, reaching density of 143 plants/ha (Cordeiro et al. 2009). The piassava palm is endemic to Bahia’s coastline. In the studied mosaic, the piassava palms are managed in agroforestry systems, where individuals in different stages of vegetative growth are randomly distributed (around 400 plants/ha), growing with other native tree species from the Atlantic forest (Guimarães and Silva 2012). The Atlantic forest covers around 10–20 % of the area of both the piassava and oil palm mosaics.

The selected forest-rubber tree mosaic encompasses the neighboring protected forest areas of Michelin Ecological Reserve (MER) with 3,096 ha of protected tropical rainforest at altitudes from 160 to 327 masl. The forest covers around 43 % of the area of this mosaic and is concentrated in two northern large blocks (625 and 550 ha, the last one connected to neighboring forests of CCC, covering more than 4,000 ha of continuous forests), in addition to narrow riverine forests (10–13 m width) and many small forest fragments (<15 ha) imbedded in rubber tree matrix which occupies about 30 % of the mosaic area (Ramalho et al. 2013). The rubber trees are often planted with 7 to 8 m spacing between-lines and 2.5–3 m in the line, with density of 476–500 trees/ha, and the adult trees can reach 40–50 m in height (Duarte 2007).

2.2 Bee sampling

In each of the three crop-forest mosaics, two odor traps were installed in three replicates of the following treatments: (1) the nuclear area of forest fragments (identified as 0 m treatment) and crop matrix in (2) 500 m, (3) 1,000 m, and (4) 2,000 m away from the edge of the forest fragment.

The two odor traps of each replicate were 100 m distant from each other and installed 1.5 m above ground. The odor trap and the attractive aromatic compound (eucalyptol or 1.8-cineol) were standardized following Ramalho et al. (2013). The field samplings were simultaneous inside each mosaic, but not among mosaics (Online Resource 2). The sampling were carried out including the most active periods of orchid bees in the region (November to March; Ramalho et al. 2013), during two consecutive days and 48 h per period, and totalizing 144 h per mosaic (in three Sampling Period, see Online Resource 2).

The sampled bee males are deposited in Pollination Ecology Lab (ECOPOL) of the Federal University of Bahia (UFBA). The species were determined by comparison with the reference material (identified by Dr. André Nemésio—Federal University of Uberlândia—MG-Brazil) and using identification keys of Nemésio (2009). Nomenclature of Nemésio (2009) was adopted. To ease communication, we utilized the following abbreviations for the genera: Eg. = Euglossa, El. = Eulaema, and Ex. = Exaerete.

2.3 Statistical analysis

First, it was made an exploratory analysis of the distance effect, grouping data on the three mosaics. Second, individual analyses were carried out for each type of mosaics by considering the following factors: distance from forest fragment and time (each sampling period was a treatment/factor level). Multivariate (species composition) and univariate (total abundance) variance analyses (MANOVA and ANOVA) were performed, the post hoc tests were made using the Tukey’s Honestly Significant Differences test, and the Pillai’s trace multivariate test method was also used. All orchid bee species in the analyses of total abundance and only the species with more than 50 sampled individuals in the analyses per species were taken into account.

The data normality and homoscedasticity were evaluated, respectively, with the Kolmogorov-Smirnov test and with the Levene’s Variances Equality test. All tests were made in the statistics program SPSS 13.0 for Windows and the significance level adopted was 0.05. The dependent variables that did not satisfy the normality premise of the analysis of variance were transformed using natural logarithm plus one (ln [X + 1], where X is the value to be transformed).

3 Results

A total of 21 orchid bee species were sampled in the three crop-forest mosaics of the AFC, from which the most abundant ones were, in descending order: Eulaema nigrita Lepeletier, 1841 (1399), Euglossa ignita Smith, 1874 (512), Euglossa carolina Nemésio, 2009 (506), Eulaema atleticana Nemésio, 2009 (486), Euglossa imperialis Cockerell, 1922 (88), and Euglossa securigera Dressler, 1982 (54) (Online Resource 3).

The species composition and the most abundant species differed among the types of mosaics, except for the Eg. carolina (Table I). In contrast, the less abundant species showed no differences between mosaics, except for the Euglossa iopoecila, which was sampled only in the mosaic of forest with rubber tree matrix (Table I, Online Resource 2).

Table I Effect of distance from the forest fragment on the permeability of arboreal matrices to Euglossina. p = probability.
Full size table

Only El. atleticana, Eg. imperialis, and Eg. iopoecila showed differences in the number of bees captured between the forest and crop matrices (Table I; Online Resource 4). The Eg. iopoecila, in particular, was collected only in the interior of forest fragments. However, the three species presented no variation in abundance associated to distances inside the three matrices (Table II).

Table II Paired comparisons (post hoc test) of the effect of “distance from the forest fragment” on the permeability of the arboreal matrices. Treatments: A = 0 m (inside the forest fragment), B = 500 m, C = 1,000 m, D = 2,000 m inside de matrix. p = probability.
Full size table

The apparent response of Eg. imperialis to the interaction of factors (mosaic type and distance) is related with its strong difference between the forest fragment and the rubber-matrix (Table III). The Eg. iopoecila significant interaction resulted of its absence in two of three mosaics (forest mosaics with oil palm and piassava palm).

Table III Effects of distance (from the forest fragment), time, and interaction of the both factors on the permeability of arboreal matrices to Euglossine bees. MRM = mosaic of forest fragments and rubber tree matrix; MPM = mosaic of forest fragments and piassava palm matrix; MOM = mosaic of forest fragments and oil palm matrix. p = probability.
Full size table

Among the most abundant species, El. atleticana and Eg. imperialis showed variation between forest and rubber tree matrix that also explain the significant spatial variation in the total abundance of Euglossina in this mosaic (Table III; Figure 1). Only El. atleticana decreased in abundance in the oil palm matrix in relation to the forest (Table III). In the piassava palm mosaic (MPM), the abundances of no species varied between the forest and the matrix and, therefore, there was no influence of the distance into the matrix in relation to the forest (Table III; Figure 2).

Figure 1.
figure 1

Spatial distribution of abundance (number of individuals) of the most abundant species of Euglossine bees in the forest-rubber tree mosaics in the Brazilian Atlantic Forest Corridor. a El. atleticana; b El. nigrita; c Eg. carolina; d Eg. ignita; e Eg. imperialis; f Eg. securigera.

Full size image
Figure 2.
figure 2

Spatial distribution of abundance (number of individuals) of the most abundant species of Euglossine bees in the forest-piassava palm mosaics in the Brazilian Atlantic Forest Corridor. a El. atleticana; b El. nigrita; c Eg. carolina; d Eg. ignita; e Eg. imperialis; f Eg. securigera.

Full size image

The time factor affected the species composition and the abundance of about 50 % of the species in the three mosaic types, and total abundance of orchid bees in the forest-rubber mosaic (Table III), that probably reflect a marked seasonal variation of flying activity of orchid bees in the region. Only Eg. imperialis responded to interaction of factors time and distance at rubber mosaic (Table III), but probably this could be due to insufficient number of individuals sampled in some period of the year (Table IV). Therefore, the spatial distribution of Euglossina in each mosaic type was weakly affected by the period of the year (Table III).

Table IV Paired comparisons (post hoc test) of the effect of factor distance from the forest fragment on the permeability of Eg. imperialis in the forest-rubber tree mosaic. Treatments: A = 0 m (forest fragment), B = 500 m, C = 1,000 m, D = 2,000 m.
Full size table

In the few cases of significant differences between forest and matrix (Table III), there were no influence of the distances in the matrix on observed variation in abundance (Table V; Figures 2 and 3). When there is some influence of the matrix on the orchid bees, this is not related to its extension in the landscape.

Table V Paired comparisons (post hoc test) of the effect of “distance from the forest fragment” on the permeability of matrices to the Euglossina, per type of mosaic. MRM, MPM, and MOM as in Table III. Treatments: A = 0 m (forest fragment), B = 500 m, C = 1,000 m, D = 2,000 m.
Full size table
Figure 3.
figure 3

Spatial distribution of abundance (number of individuals) of the most abundant species of Euglossine in the forest-oil palm mosaics in the Brazilian Atlantic Forest Corridor. a El. atleticana; b El. nigrita; c Eg. carolina; d Eg. ignita; e Eg. imperialis; f Eg. securigera.

Full size image

4 Discussion

In general, the palms and rubber tree matrices were permeable to most of the Euglossina species. The piassava palm crop was more heterogeneous and more permeable than the other two matrices, as expected. Some studies have pointed out that the heterogeneity and quality of matrices affect the richness and abundance of bees and wasps in general (both from the Hymenoptera group), with varied influences on different species (e.g., Steffan-Dewenter and Tscharntke 1999; Steffan-Dewenter 2003; Jauker et al. 2009).

The piassava is harvested in areas of native forest submitted to extensive selective cut and clearing of large native trees. This agroforestry system still preserves native plant and mainly understory species (Guimarães and Silva 2012) that offer key flower resources to bees’ preservation (Roulston and Goodell 2011), moderate shading conditions, and therefore, better general conditions of occupation by orchid bees.

Of the 21 sampled Euglossina species, only three have reduced abundances in the matrices: El. atleticana in the oil palm tree matrix; El. atleticana, Eg. imperialis, and Eg. iopoecila in the rubber tree matrix. This last orchid bee was sampled exclusively in the more extensive forest block of the region, indicating sensitivity to forest fragmentation and disturbance. The reduction of permeability to the other two Euglossina species did not lead to the loss of functional connectivity in the mosaic scale, at least to the level of current forest fragmentation.

El. atleticana and Eg. imperialis were the only ones who responded to the forest fragment size and presented lower capture rate in rubber tree crops and other anthropic habitats compared to the nearby forest areas (Milet-Pinheiro and Schlindwein 2005; Aguiar 2011; Ramalho et al. 2013; this study). Therefore, evidences indicate that these two species are presenting lower ability of exploiting the anthropogenic habitats in expansion with regional deforestation (this study; Aguiar 2011; Ramalho et al. 2013). Although they are able to explore non-arboreal environments, they seem to be sensitive to changes of forest cover in the landscape. Therefore, El. atleticana and Eg. imperialis are good candidates to function as bioindicators.

On the other extreme, the species El. nigrita and Eg. carolina were weakly affected by fragmentation metrics, such as the forest fragment size (e.g., Aguiar 2011; Ramalho et al. 2013), and they forage in different matrices, moving through extensive areas of open vegetation (e.g., Tonhasca et al. 2003; Aguiar 2011; Ramalho et al. 2013). Because of these characteristics, they could mitigate the rupture of important ecosystem functions in extremely deforested landscapes of AFC, in particular the long-distance pollination.

There is no consensus regarding the intensities and main effects of the forest fragmentation on the Euglossine bees (see brief review in Ramalho et al. 2013). However, most (70 %) of the studies have reported effects of some fragmentation metrics (for example, fragment size or core area, area-edge ratio, habitat quality) on the diversity and/or the abundance of those bees (e.g., Powell and Powell 1987; Raw 1989; Peruquetti et al. 1999; Otero and Sandino 2003; Sofia and Suzuki 2004; Darrault et al. 2006; Brosi 2009; Ramalho et al. 2009, 2013; Nemésio and Silveira 2010; Aguiar and Gaglianone 2012). On the contrary, few cases did not report any effects of the forest fragment size or quality on orchid bees (Becker et al. 1991; Tonhasca et al. 2002; Rasmussen 2009).

Therefore, the Euglossine bees seem to be affected by the forest fragmentation to the habitat patch level, and often their abundance and/or richness respond to the area, shape, and/or quality of fragment. This kind of relationship is highly expected and had already been observed in various taxonomic groups since the theory of islands biogeography by MacArthur and Wilson’s (1967) became popular in the studies on habitat fragmentation (Kattan and Murcia 2003).

Some studies point out the high correlation between the body size and foraging distance in different groups of bees (e.g., Van Nieuwstadt and Ruano Iraheta 1996; Gathmann and Tscharntke 2002; Greenleaf et al. 2007). Most Euglossines are large bees and various species stand out for its very large body size, for example, El. atleticana and El. nigrita have average body size of 26.7 and 19 mm, respectively (Nemésio 2009). Thereby, the size of some orchid bees offers an explanatory component to long forage distances or willingness of crossing or exploring non-forested environments (e.g., Janzen 1971; Powell and Powell 1987; Roubik and Hanson 2004; Wikelski et al. 2010).

However, in this study, there is no evidence that the size has relevant influence or predictive power on the responses of orchid bees, e.g., the observed similar responses to matrices of two species with very different body sizes: El. atleticana and Eg. imperialis (26.7 and 15 mm, respectively; Nemésio 2009).

With regard to mobility of orchid bees, the relevant issue should be the quality of the matrices in terms of floral resources (e.g., Roulston and Goodell 2011), but not necessarily in terms of the physiognomy (e.g., arboreal and non-arboreal matrices; see also Ramalho et al. 2013). If the density of planted trees/ha would have relevant influence on the quality of the agricultural matrices to the orchid bees, the rubber matrix (500 trees/ha) should provide greater permeability than the oil palm tree (140 palm trees/ha), and this was not confirmed. The effects of the quality and heterogeneity of the matrices on mobility were reported on various groups of insects (e.g., Ricketts 2001; Winfree et al. 2011).

Regarding most of the Euglossina bee species, the arboreal matrices of piassava palm, oil palm, and rubber trees must contribute to maintain a moderate connectivity level and gene flow (Rosa et al., unpublished data), considering the current level of forest cover and fragmentation in the northern AFC. However, the species-specific responses are becoming evident, considering the differential sensibility of El. atleticana and Eg. imperialis to the current fragmentation level.

The northern portion of the Atlantic Forest Corridor is composed by a high proportion of small forest fragments (<10 ha) and riverine forests, and rare large fragments (>100 ha) (Fundação SOS Mata Atlântica and INPE 2013; Ramalho et al. 2013). The small fragments and the riverine forests are still being occupied by Euglossine bees (Ramalho et al. 2013) and might be serving, respectively, as stepping stones or biological corridors in the current stage of forest fragmentation. However, with increasing risk of disappearance of such connecting elements at the landscape scale, the regional arboreal matrices with depleted understory vegetation must loose connectivity to most sensitive species of orchid bees living in forest.