Otero And Sandino 2003

BIOTROPICA 35(4): 520–529

2003

Capture Rates of Male Euglossine Bees across a Human Intervention Gradient, Choco´ Region, Colombia1 J. Tupac Otero 2,3 Departamento de Biologı´a, Universidad del Valle, A.A. 25360, Cali, Colombia and Juan Carlos Sandino 4 Fundacio´n Herencia Verde, A.A. 38302, Cali, Colombia

ABSTRACT Euglossine bees are important pollinators of lowland Neotropical forests. Compared to disturbed habitats, undisturbed ones have been previously characterized by higher abundance and diversity of euglossine bees. Most past studies have relied on chemically baiting male bees at single sites within habitats. Over a two-year period, we employed a repeatedmeasures design in which we sampled bees at multiple sites within three different habitat types, reflecting a mosaic of human disturbance (farm, secondary forest, and old logged forest). After 22 monthly samples, a total of 2008 male bees were captured, representing 31 species in five genera: 1156 at the farm (57.6%, 21 spp.), 505 in the secondary forest (25.1%, 27 spp.), and 347 in the old logged forest (17.2%, 21 spp.). Eighty-one percent of the bees captured belonged to the five most abundant species: Eulaema cingulata, El. chocoana, Euglossa hansoni, Eg. ignita, and Eg. imperialis. These species differed significantly in capture frequencies among habitats. Eulaema cingulata, El. chocoana, and Eg. ignita were captured most frequently at the farm, while Eg. imperialis was most abundant in the secondary forest. In contrast, Eg. hansoni, the sole short-tongued species among the five, was equally abundant in the two forest habitats but occurred rarely on the farm. Additionally, habitats differed in bee composition. The high capture rates for long-proboscis species at the farm may have been due to their ability to extract nectar from flowers with long floral tubes, which probably occurred at a greater density on the farmed land than in the adjacent forests.

RESUMEN Las abejas euglosinas son polinizadores importantes en bosques neotropicales de tierras bajas. Normalmente, los ha´bitats poco perturbados se caracterizan por presentar una mayor abundancia de abejas euglosinas que los menos intervenidos, aunque dichos resultados se basan en muestreos que usan atrayentes quı´micos para machos en sitios u´nicos por ha´bitat. En este estudio realizamos muestreos basados en sitios mu´ltiples en tres ha´bitats adyacentes (finca, bosque secundario, y bosque maduro) a lo largo de dos an˜os. Encontramos una mayor abundancia y diversidad de machos euglosinos en los ha´bitats con mayor perturbacio´n humana que en el bosque maduro. Despue´s de 22 muestreos mensuales capturamos un total de 2008 abejas representando 31 especies en cinco ge´neros: 1156 en la finca (57.6%, 21 spp), 505 en el bosque secundario (25.1%, 27 spp), y 347 en el bosque maduro (17.2%, 21 spp). El 80.8 por ciento de las abejas capturadas pertenecı´an a las cinco especies ma´s abundantes: Eulaema cingulata, El. chocoana, Euglossa hansoni, Eg. ignita, y Eg. imperialis. Para estas especies encontramos diferencias en la frecuencia de captura entre ha´bitats. Eulaema cingulata, El. chocoana, y Eg. ignita fueron ma´s frecuentes en la finca, mientras que Eg. hansoni, la u´nica de las cinco con lengua corta, lo fue en los dos ha´bitats boscosos y Eg. imperialis en el bosque secundario. Adicionalmente, los ha´bitats difirieron en composicio´n de abejas. Lo resultados pueden deberse a que en la finca habı´a una mayor oferta de ne´ctar con acceso restringido que en los busques adyacentes. El ne´ctar, por estar en flores de corolas profundas, solo podı´a ser accedido por abejas de lenguas largas lo cual favorecı´a una mayor abundancia de euglosinas grandes con lenguas largas en la finca que en los bosques adyacentes. Key words:

Choco´; Colombia; Euglossa; euglossine bees; Eulaema; human intervention gradient; pollinators.

EUGLOSSINE BEES (EUGLOSSINI, APIDAE) ARE AMONG the most important long-distance pollinators of 1 Received 18 November 2002; revision accepted 27 August 2003. 2 Corresponding author; e-mail: [email protected] 3 Current address: CSIRO Plant Industry, Australian National Herbarium, GPO Box 1600, Canberra, ACT 2601, Australia. 4 Current address: Facultad de Biologı´a Marina, Universidad Jorge Tadeo Lozano, Carrera 22-61, Bogota´, Colombia.

lowland Neotropical forests (Bawa 1990). The ca 200 species of this group (Kimsey & Dressler 1986) pollinate a vast array of plants at all successional stages (Gilbert 1980, Dressler 1982a, Ackerman 1985). The taxonomy, ecology, and natural history of euglossine bees are well documented (Zucchi et al. 1969, Ackerman et al. 1982, Dressler 1982a, Janzen et al. 1982, Williams 1982, Roubik & Ackerman 1987, Roubik 1989, Bonilla & Nates 1992, Armbruster 1993, Ferna´ndez 1995). Their most remarkable feature is the fragrance-foraging

520

Euglossines across an Intervention Gradient

activity of males at flowers of several different plant families and non-floral sources such as rotting logs (Williams 1982, Whitten et al. 1993). The role of euglossine bees as long-distance pollinators is critical because many plant species require cross-pollination (Bawa 1990, Kress & Beach 1992, Oyama 1993) but exist at very low population densities (Faber-Langendoen & Gentry 1991, Clark 1992). Some authors have hypothesized that patches of abundant resources resulting from deforestation and farming practices may disrupt the foraging dynamics of euglossine bees in adjacent forests, and thus the gene flow they mediate as pollinators may be severely affected (Janzen 1974, Aldrich & Hamrick 1998). Since identifying the main components of the attractive fragrances, commercially available chemicals have been used for obtaining, in a few days, large samples of euglossine communities at given localities (Janzen et al. 1982, Ackerman 1983, Pearson & Dressler 1985, Powell & Powell 1987, Roubik & Ackerman 1987, Sandino 1995a). Most such samplings have been done at single locations within habitats and at single habitats within landscapes. Capture rates are assumed to reflect actual bee abundance at each habitat (Roubik 2001). Because foraging for nectar and fragrances and sexual displays may each be performed at different and distant sites (Janzen et al. 1982), we can assume that capture rates reflect the use that males make of each habitat. Some evidence supports this assumption. Male euglossine bees may be collected in higher frequency near the sites from which they obtain food or fragrances (Ackerman 1983, Armbruster 1993). Because they do not collect fragrances everyday, but do feed frequently, it seems logical to suppose nectar distribution as a main, but not sole, causal factor of the capture frequencies. And because euglossine bees probably forage across ample distances between patches (Janzen 1971, 1981), and given evidence of between-site variability in bait–capture samples from a single habitat (Ackerman 1983, Armbruster 1993; cf. Roubik 2001), euglossine foraging and demographic dynamics may be better addressed by long-term sampling and multiple-habitat and multiple-site sampling. Previous studies on the effects of deforestation and fragmentation on euglossine bee communities have been incomplete (Cane 2001) and based on single-site sampling (Powell & Powell 1987, Becker et al. 1991). The main objective of our study was to document the differences in euglossine bee community structure through a human intervention

521

gradient at a lowland site in the Choco´ biogeographic region of Colombia. To accomplish this goal, we employed a repeated-measures design in which, over a two-year period, we followed specific sampling plots in three habitat types reflecting different levels of human disturbance. Specifically, we asked: (1) Are euglossine bees more abundant and diverse in forested habitats than in a farm?; (2) Do capture rates differ among sites within habitats?; (3) Do forested and farm sites differ in euglossine bee composition?; (4) What bee characteristics may be related to capture rates among the sampled habitats?; and (5) Do capture rates vary across years?

METHODS STUDY SITE. This study was conducted in the central Pacific plain of Colombia by the old Cali-Buenaventura road (completed in 1946, but now seldomly used), which is near the town of Guaimı´a (38469N, 768579W) on the Anchicaya´ River. The area, part of the biogeographical region of Choco´, receives more than 7000 mm of annual rainfall and has an average relative humidity of 96 percent. Study sites were between 50 and 80 m elevation, ranging from the riverbanks to low hills. The forest is classified in the Holdridge system as transitional between tropical wet and tropical pluvial forest (IGAC 1977). Indigenous peoples first settled in this zone at least 2200 years ago (Herrera 1989) and were replaced by Afro–American communities that followed similar farming practices. The present human community and landscape have resulted from the merging of traditional self-sufficient communities of farmers and miners with more recent colonizers. The present Afro–Colombian community depends mainly on a subsistence economy based on timber extraction, mining, agriculture, fishing, and hunting. Farmlands are present all along the Anchicaya´ River, and secondary forests abut the farms and roads. Based on their accessibility, we chose three different habitats that represented the most common habitat gradients within the landscape (Tables 1 and 2). These habitats were a traditional farm (Fa), a selectively and intensively logged secondary forest (SFo), and a less disturbed old logged forest (OFo). The farm, named Limones and property of the community organization AFEPAL, was a multicrop farm having several cultivars that included both food and timber products (Table 1). It was 80 ha on the west margin of the Anchicaya´ River and adjacent to secondary forests. The secondary

522

Otero and Sandino

TABLE 1. Plant species at the study sites (R. Ospina, pers. comm.). Site Farm

Secondary forest

Old logged forest

Plant species Cultivars: Chontaduro (Bactris sp.), green plantain (Musa paradisiaca), cassava (Manihot esculenta), breadfruit (Artocarpus altilis), sugarcane (Saccharum officinale), and borojo´ (Borojoa patinoi) Dominant trees: Cedrela odorata (Meliaceae) and Cecropia spp. (Cecropiaceae) Herbs reported as euglossine resources: Calathea spp., Heliconia spp., and Anthurium spp. Dominant trees: Mabea sp. (Euphorbiaceae), Cespedesia spatulata (Ochnaceae), Otoba latialata (Myristicaeae), Miconia spp. (Melastomataceae), Wetinia quinara (Arecaceae), Chrysophillum sp. (Sapotaceae), Brosium utile (Moraceae) Herbs reported as euglossine resources: Anthurium spp. Dominant trees: Dussia lehemani (Fabaceae), Socratea exorrhiza (Arecaceae), Licania duriflora (Chrysobalanaceae), Cecropia sp. (Cecropiaceae), Brosium utile (Moraceae), and Symphonia globulifera (Clusiaceae) Herbs reported as euglossine resources: Anthurium sp. and other plant species

forest had constant human intervention for timber extraction and hunting by local people. The old logged forest had intense commercial timber extraction from the 1940s until ca 15 years before our study, although selective timber extraction along the Yesqueros stream was still frequent for local uses. The old logged forest, however, was continuous with a natural undisturbed forest that is part of the Buenaventura municipal reserve of San Cipriano. In each habitat, we established four sampling sites (Fa 1–4, SFo 5–8, and OFo 9–12) located at least 200 m away from each other, in agreement with Armbruster’s (1993) heterogeneous results between sites. HABITAT STRUCTURE. At each of four sampling sites per habitat, we estimated the canopy height using a clinometer, the average diameter of trees at breast height (DBH), and the tree density as the number of trees larger than 10 cm DBH within a 40 3 40 m plot. During the last months of our study, a floristic profile was determined by other researchers along our sampling routes. Their specimens were deposited at the Herbarium of the Universidad del Valle (CUVC). BEE SAMPLING DESIGN. At each sampling site, we placed three homemade, nonlethal traps (Sandino

in press) made from plastic dispensable bottles and nylon stockings attached to the top in which the bees arrived looking for the fragrance and were added to a vial. The bees were able to enter the trap to collect the fragrance, but once inside were trapped in the stockings when attempting to leave (Sandino 1995a). We used three different chemical baits, one per trap at each site: 1-8 cineole, methyl salycilate, and skatole. Bees were active between 0730 and 1430 h (Sandino, pers. obs.). Traps were opened from 0800 until 1200 h and checked every 30 minutes. The three habitats were sampled simultaneously by three different collectors. Samples were collected every four weeks, when possible, from 25 June 1995 to 28 July 1997. Bees captured were identified in the field using a guide based on a previous reference collection from the same locality and also by consulting specimens available from other studies in the Choco´ region (Sandino 1995b) and elsewhere in Colombia (Bonilla & Nates 1992). If a bee could not be identified in the field, it was collected. Voucher specimens were deposited in the bee laboratory at the Universidad Nacional, Bogota´, Colombia, and were identified using keys from the literature (Dressler 1978a, b, 1979, 1982b; Kimsey 1982; Bonilla & Nates 1992; Ospina 1998) and by comparison with bees deposited there. Once identified,

TABLE 2. Structural measurements of the study sites (means 6 SD).

Site

Canopy height m (6SD)

Trees .10 cm DBH/ha no. (6SD)

Average DBH cm (6SD)

Farm Secondary forest Old logged forest

12.4 (66.2) 13.2 (64.4) 20.7 (66.5)

5875 (63025) 19,925 (64225) 7625 (61625)

50.1 (656.7) 27.3 (621.1) 44.4 (639.2)

Euglossines across an Intervention Gradient

bees were marked on a wing using an indelible ink marker (Outliner, Sakuraq, Hayward, California) with a color specific to each habitat and then set free. In marking bees, our goal was to estimate recapture frequencies to assess the independence of our sampling regime as well as to estimate population sizes if recaptures were high. MORPHOMETRIC RANKS OF THE BEES. Even folded, the proboscis is notably long in many euglossine species (thus the name for the taxon), in some cases exceeding body length. Species with short-folded probosces are generally those Euglossa in which the proboscis does not reach the abdomen. Euglossa bees are more like honeybees in size, while Eufriesea and Eulaema are more bumblebee-like. Thus, we grouped the vouchers according to the above distictions. Because wing length is linearly correlated with body size (Kimsey 1982) and tongue measurements vary 2–10 percent (Roubik 1992) within species, we used the folded proboscis length and the anterior wing length of bees to classify each species in one of three morphometric classes: large bees, with wings larger than 1.2 cm and folded proboscis longer than 0.8 cm; medium-sized, longtongued bees, with wings shorter than 1.2 cm and folded proboscis longer than 0.8 cm; and mediumsized, short-tongued bees, with wings shorter than 1.2 and folded proboscis shorter than 0.8 cm. DATA ANALYSIS. To test for differentiation in habitat structure, we performed a Kruskal–Wallis test (KW) on the canopy height and the tree density data. With the 22 samples of bees, we created frequency tables for each of the 12 sites (264 cases). We tested bee frequency distribution for normality, using the Anderson–Darling test. We calculated the abundance, richness, the Shannon–Wiener diversity index (H9), and the evenness (J9) according to Zar (1999), and the Simpson’s index of dominance (li) that measures the probability of randomly selecting two individuals of the same species in a community (Brower et al. 1997). To test if species richness and abundance of bees were equal across all habitats, we performed repeated measures ANOVA (Zar 1999) on the natural log transformed data for number of bees and the number of bee species. We repeated these analyses using abundance data from the five most abundant species, because those corresponded to 80.8 percent of the captures. Using the repeated measures ANOVA design, we tested for differences among sites within habitats (subject effects), differences among habitat types (treatment effects), and

523

the interaction between habitat type and time. Because we sampled over two years, our time effect (22 mo) represented seasonal effects and temporal dynamics. To determine whether or not the bee community changed from the first to the second year, two principal components analyses (PCA) were conducted using a capture frequency matrix of 31 species 3 12 sites including the bees captured during both the first and second year of the study. To determine whether or not there was an overall differentiation in bee composition among habitats, we performed a combined PCA with a matrix including the samples from both years. To compare the 12 communities for year 1 versus year 2, a PCA was done with a matrix that included samples from each year (June 1995–June 1996 vs. July 1996– July 1997). All PCAs were performed using the procedure PRINCOMP (SAS 1989). Finally, to test differences among habitats, we used a stepwise discriminant analysis using the 13 most abundant species (PROC DISCIM; SAS 1989). Significant variables (bee species) that remained in the model were then used in a canonical discriminant analysis (PROC DISCIM; SAS 1989).

RESULTS STRUCTURAL ANALYSIS OF THE HABITATS. Between habitats, there were significant differences in tree density (Kruskal–Wallis, H 5 8.00, df 5 2, P 5 0.018) and in the average DBH (H 5 25.12, df 5 2, P , 0.0001) but not in canopy height (H 5 3.58, df 5 2, P 5 0.167). At the farm, the canopy averaged 12.4 m (6SD 6.2) and tree density was 23.5 (6SD 12.1) per 1600 m2. The secondary forest was the most homogeneous of the habitats in terms of the DBH classes and canopy height (13.2 m 6SD 4.4) and had the highest tree density (79.7 6SD 16.9 trees/1600 m2). The old logged forest had the tallest canopy, 20.7 m (6SD 6.5), but relatively low tree density (30.5 6SD 6.5 trees/1600 m2). BEE ABUNDANCE AND DIVERSITY. Over the 22 sampling dates, we collected 2008 male euglossine bees belonging to 31 species and five genera (Table 3). No marked bees were recaptured. At the farm sites, we captured 1156 males (57.6% of the total captures) belonging to 21 species; at the secondary forest sites, 505 males (25.1%) of 27 species; and in the old logged forest sites, 347 males (17.2%) of 21 species. The number of captured bees was significantly different across habitats (F2, 9 5 33.02, P , 0.0001). More bees were captured on the farm

524

Otero and Sandino

TABLE 3. Total captures of male euglossine bees in each of the three study habitats. Fa: farm; SFo: secondary forest; and OFo: old logged forest. The three morphometric ranks are L/L: large bees (with wings larger than 1.2 cm and proboscis longer than 0.8 cm); M/L: medium-sized with long proboscis (with wings shorter than 1.2 cm and proboscis longer than 0.8 cm); and M/S: medium-sized with short proboscis (with wings shorter than 1.2 and proboscis shorter than 0.8 cm). li: Simpson’s index of dominance. H9: Shannon–Wiener index. J9: evenness index. Species Eulaema aff. bombiformis El. chocoana El. cingulata El. sororia Euglossa allosticta Eg. asarophora Eg. bursigera Eg. chalybeata Eg. championi Eg. deceptrix Eg. dressleri Eg. flammea Eg. gorgonensis Eg. hansoni Eg. ignita Eg. imperialis Eg. aff. laevecincta Eg. mixta Eg. townsendi Eg. tridentata Euglossa sp. 9 Glossurela sp. 1 Glossurela sp. 9 sp. T 21 sp. T 23 sp. Q 24 Exaerete frontalis Ex. smaragdina Eufriesea pulchra Eufriesea sp. Aglae caerulea Total Bees Species li H9 J9

Morphometric rank L/L L/L L/L L/L M/L M/L M/L M/L M/S M/S M/S M/L M/L M/S M/L M/L M/L M/S M/S M/S M/S M/L M/L M/L M/L M/L L/L L/L L/L L/L L/L

than in both the secondary forest and the old logged forest, but there were no differences between the number of captures in the secondary forest and the old logged forest (Tukey post hoc test). For three of the five most frequently collected species, there were significant differences in the number of bees captured among habitats (Fig. 1): Eulaema cingulata, (F2, 9 5 53.96, P , 0.0001), El. chocoana (F2, 9 5 10.56, P , 0.0043), and Eg. ignita (F2, 63 5 5.56, P 5 0.0267). The other two species, Euglossa hansoni (F2, 9 5 1.69, P 5 0.2379) and Eg. imperialis (F2, 9 5 3.30, P , 0.0842), showed nonsignificant differences across habitats.

Fa 24 294 329 7 27 2 12 4 3 63 7 33 276 17 2 8 14 17 14

SFo

OFo

Total

20 27 38 5 6 1 9 2 16 1 1 4

2 3 50 2 6 8 2 8 18

46 324 417 14 39 9 13 10 46 5 11 77 11 300 437 145 1 1 21 10 16 20 19 1 1 1 1 6 4 1 1

164 76 107 1 1 10 2 1 5 1 1

7 10 4 103 85 21 9 2 2

1 1 2

1156 21 0.21 0.84 0.63

3 1 1 1

3 1

505 27 0.18 0.92 0.64

347 21 0.18 0.93 0.70

2008 21 0.15 0.97 0.66

These five species accounted for 80.8 percent of the captured males, none of which were more abundant in the old logged forest than in other habitats. Eulaema chocoana, El. cingulata, and Eg. ignita were significantly more frequent at the farm than at the other sites (X2 5 148.49, X2 5 82.32, and X2 5 14.02, respectively; all tests df 5 2, P , 0.005), while Eg. hansoni and Eg. imperialis were captured most frequently in the secondary forest (X2 5 266.70 and X2 5 109.07, respectively; both with df 5 2, P , 0.005). The number of species captured per habitat varied significantly (F2, 9 5 13.78, P , 0.0018). There was no significant dif-

Euglossines across an Intervention Gradient

525

FIGURE 1. Number of captures for the five most common species through the human intervention gradient. Fa: farm; SFo: secondary forest; and OFo: old logged forest.

ference in the H9 index (Kruskal–Wallis, H 5 1.38, df 5 2, P 5 0.50) and evenness values J9 (H 5 2.28, df 5 2, P 5 0.31) for the different habitats (Table 1). BEE COMPOSITION. The first principal component of the overall PCA based on both years explained 26 percent of the variability and separated farm and forested sites (Fig. 2c). This component was positively correlated with the abundance of El. chocoana, El. cingulata, and Eg. flammea and negatively correlated with Eg. hansoni. The second principal component explained 20 percent of the variability and separated secondary forest sites from old forest sites (Fig. 2c). Additionally, it separated one of the farm sites (2) from the other three sites (1, 3, and 4). The second principal component was positively correlated with the abundance of Eg. ignita and Eg. chalybeata, and negatively correlated with El. bombiformis and El. chocoana. Stepwise discriminant analysis for the habitat sites showed that four bee species contributed to habitat differentiation (El. bombiformis, F3, 11 5 5.22, P 5 0.0410; El. chocoana, F3, 11 5 9.75, P 5 0.0056; El. cingulata F3, 11 5 6.08, P 5 0.0361; and Eg. flammea F3, 11 5 14.56, P 5 0.0022). The discriminant function based on the four bees as predictor variables collectively resulted in a perfect (100%) classification of sites among habitats. The difference between farm and forested sites was consistent for both study years (Fig. 2a, b). In the first year PCA, the first principal component explained 32 percent of the variability and showed a differentiation between the farm and forested sites. The second principal component explained 22 percent of the variability. This component showed a differentiation between secondary forest and old forest sites (Fig. 2a). In the second year,

FIGURE 2. Plot of the two principal components of the PCA based on the proportion of bees captured at the 12 sites during the (a) first year, (b) second year, and (c) combined first and second years. Sites Fa-1–Fa-4 are on the farm (Fa); Sf-5–Sf-8 in the secondary forest (SFo); and Of-9–Of-12 in the old logged forest (OFo).

526

Otero and Sandino

the first principal component explained 28 percent of the variability and showed a differentiation between the farm and forested sites as in the first year. The second principal component explained 21 percent of the variability. This component also showed a differentiation between three secondary forest sites and the other secondary forest site plus old forest sites (Fig. 2b). MORPHOMETRIC RANKS. Sorted on morphometric ranks, bee captures were uneven both across and within habitats. The proportion of large bees (H 5 6.27, df 5 2, P 5 0.04) and medium-sized, short-tongued bees (H 5 8.0, df 5 2, P 5 0.02) differed significantly among habitats. Large bees and medium, long-tongued bees overall were more frequently captured at the farm (F2, 9 5 22.59, P 5 0.0003 and F2, 9 5 4.05, P 5 0.0441, respectively), while the abundance of medium, shorttongued bees did not vary among habitats (F2, 9 5 1.15, P 5 0.3592). On the farm, most of the captured bees were large (57.1% of the bees, represented by 6 spp.) and long-tongued (93.1% of the bees, represented by 14 spp.). Medium-sized, shorttongued bees were uncommon (6.9% of the bees). In both the secondary and the old logged forests, large bees were less abundant (19 and 17.6% of the bees, respectively), as medium-sized, shorttongued bees were more frequent (39.0 and 40.3% of the bees, respectively). In contrast, there were no significant differences in the proportion of medium-sized, long-tongued bees among habitats (H 5 1.50, df 5 2, P 5 0.47). TEMPORAL VARIATION IN BEE ABUNDANCE. There was temporal variation in the abundance of euglossine bees over the two study years (F4, 36 5 5.76, P 5 0.0010; Fig. 3). This variation was partially due to abundance peaks associated with high capture rates for some species. There was a general increase in the number of captured bees in the last six months of the study. More bees were captured during the last six months compared to the previous period (F1, 5 5 13.3, P 5 0.015). Nevertheless, the effect of time on the number of species was only marginally significant (F6, 54 5 2.12, P 5 0.0654). For two of the five most frequently collected species, there were significant differences in the number of bees captured among sampling dates. Eulaema cingulata (Univar G–G Epsilon 5 0.13; F3, 25 5 9.57, P 5 0.0003) and Eg. ignita (Univar G–G Epsilon 5 0.14; F3, 26 5 4.40, P 5 0.0137) showed significant temporal variation. Large (Uni-

FIGURE 3. tures.

Temporal variation in euglossine bee cap-

var G–G Epsilon 5 0.16; F3, 30 5 7.84, P 5 0.0004) and medium-sized, long-tongued bees (Univar G–G Epsilon 5 0.17; F3, 32 5 2.78, P , 0.0481) varied temporally, while small, shorttongued bees (Univar G–G Epsilon 5 0.08; F2, 14 5 2.02, P 5 0.1729) did not.

DISCUSSION Our analyses showed that the three habitats were distinctly different in bee composition, with greater differences between farm and forest habitats than between the two forest habitats. Overall, our study confirmed, at a larger temporal scale, previous results from a two-month study at the same sites (Sandino in press): male euglossine bees were more frequently captured at farm sites than in either forested site. Capture abundances for some species changed seasonally and site samples changed from one year to another, but the pattern of more abundant, richer, distinct captures at the farm sites prevailed. We emphasize that we collected data from a single human intervention gradient, and so our results may not be applicable to other localities.

Euglossines across an Intervention Gradient

Our results were consistent with two other recent studies of euglossine bee abundance. First, a four-month study had sampled 13 euglossine species from a logged and silviculturally treated plots in Costa Rica (Rinco´n et al. 1999). They found that in the logged plots, the original old-growth forest was lightly logged at irregular intervals during 1960– 1989, while in the silviculturally treated plots, commercial logs were harvested at 10 m3/ha in 1989 and 1990. The more disturbed silviculturally treated plots had more euglossine bees visiting flowers than the logged plots. Second, a study had been conducted in Brazilian Atlantic Forest with a fragmented landscape (Tonhasca et al. 2002). When we performed a PCA analysis on their euglossine bee data, the species composition of that study revealed patterns similar to those found in our study. The first principal component differentiated fragmented from contiguous forest and the second principal component differentiated secondary from more preserved forest (disturbed but adjacent to the main forest). A study in the Brazilian Amazon, however, showed the opposite effect; there were fewer bees (species and individuals) in experimentally deforested areas than in the adjacent forest (Morato 1994). That study compared pristine forest with a pasture on the border of the forest, and the captures in the pasture may have reflected the fact that bees live in the forest and are able to forage in pasture. The differences between those studies may have been due to differences in the level of intervention among the studied sites, since the Amazonian forest (Morato 1994) was more heavily disturbed than both the Brazilian Atlantic Forest and the logged forest in Costa Rica. Based on the assumption that male euglossine capture rates reflect actual abundance at the sampled habitats, we propose that there is a morphometric trend with large-sized, long-tongued bees and medium-sized, long-tongued bees foraging more at the farm sites and medium-sized, short-tongued bees foraging more at secondary and mature forest sites. We hypothesize that the morphometric trend is correlated with two factors: the effect of food resources on proboscis length and the effect of microclimate on body size. On one hand, a long proboscis allows access to nectar from an ample range of flowers, including the long tubular flowers of Costaceae, Heliconicaeae, Marantaceae, and Apocynaceae, which were abundant in the disturbed farm habitat. The low frequency of short-tongued bees on the farm may indicate that this habitat is very competitive and that species are excluded because they do not have access to nectar from deep-corolla flowers. Additionally, large size permits bees to tolerate high

527

temperatures and low humidity (May & Casey 1983) that occur in the canopy (Roubik 1993; cf. Otero & Sallenave in press), which are expected in highly disturbed habitats like the farm with its low canopy coverage and wide cleared patches. Large species and some medium-sized (but long-tongued) species could thus be considered opportunistic, as they seem better adapted to compete for the floral resources abundant in these disturbed habitats. Nonetheless, it should be noted that our sampling included only four of the seven hours that euglossine bees are active; thus, captures may have been biased toward large bees that forage earlier in the day (D. Roubik, pers. comm.). On the other hand, shorttongued bees may be more efficient at extracting nectar from shallow flowers than long tongued-bees. Also, results from one other study seem to contradict ours. In Brazil, larger Eulaema species were found to be less tolerant of open spaces than Euglossa species (Peruquetti et al. 1999); thus, physiological traits related to tolerance of disturbance may also influence the most common species at each habitat. Data on vegetation structure gave us no further insight into habitat characteristics that contributed to male euglossine capture patterns, but as implied above, floral composition deserves more detailed study. Our results for El. chocoana are intriguing. This recently discovered species (Ospina & Sandino 1997), despite being very abundant in a disturbed habitat, appears to have a very limited distribution, apparently being endemic to the south of the Choco´ biogeographic region (Sandino 1995b). The also recently described El. sororia (Dressler & Ospina 1997), however, is abundant south of the Choco´ region (Sandino 1995a) and has been collected over a broader area, but was rarely collected in Anchicaya´ (Table 1). Why does El. chocoana, a bee species so frequently captured in disturbed environments, have such a restricted distribution? In what kind of natural habitats did the species occur before human disturbance patches were created in the region? Overall, our results show that some widely distributed (and the endemic El. chocoana), large bodied and/or long-tongued opportunistic species seem to occur more frequently in diversified farmlands and disturbed forests within a mosaic that includes old logged forest. Could this mean that previously ample foraging routes of male euglossine have decreased in farmlands, as some have suggested (Janzen 1974, Aldrich & Hamrick 1998), or that opportunistic species simply flourish at disturbed but resource-rich habitats? Our results could be interpreted in both ways. Similar multisite sampling schemes that simultaneously compare forests with

528

Otero and Sandino

adjacent monocultivars or non-diversified farmlands are needed to fully understand the effects of human land use on forest pollination dynamics.

ACKNOWLEDGMENTS We thank P. Chacon and P. Silverstone-Sopkins for their mentoring, support, and tolerance. Rodulfo Ospina, G. Nates, and R. Dressler provided taxonomical advice and facilities. Roman Ospina made the unpublished results of the botanical survey in the area available. Jim Ackerman, P. Bayman, A. Sabat, N. Flanagan, D. Roubik, R. J. Mar-

quis, and anonymous reviewers commented on early versions of the manuscript. T. A. Crowl and P. Thrall provided statistical advice. This study was possible due to the data collection and preliminary analyses made by M. Santaella, D. Dı´az, C. Restrepo, H. Asencio, A. Useche, M. I. Vallejo, O. Castro, C. Renterı´a, and I. Restrepo. Miller, William, Tocayo, and F. Angulo gave us their expert field guidance and assistance. We especially thank M. Angulo and the family of P. Angulo for their kindness and support. Permits were provided by AFEPAL and the community of Guaimia. This research was partially funded by a research grant given by the Fondo FEN-Colombia to the first author and was supported under special arrangements and contracts by the Fundacio´n Herencia Verde.

LITERATURE CITED ACKERMAN, J. D. 1983. Diversity and seasonality of male euglossine bees (Hymenoptera, Apidae) in central Panama. Ecology 64: 274–293. ———. 1985. Euglossine bees and their nectar hosts. In W. G. D’Arcy and M. D. Correa (Eds.). The flora and natural history of Panama, pp. 225–233. Missouri Botanical Garden, St. Louis, Missouri. ———, M. R. MESSLER, K. L. LU, AND A. M. MONTALVO. 1982. Food foraging behaviour of male euglossine bees: Vagabonds or trapliners? Biotropica 14: 241–248. AIZEN, M., AND P. FEISINGER. 1994. Habitat fragmentation, native insect pollination, and feral honey bees in Argentine ‘‘chaco Serrano’’. Ecol. Appl. 4: 378–392. ALDRICH, P. R., AND J. L. HAMRICK. 1998. Reproductive dominance of pasture trees in a fragmented tropical forest mosaic. Science 281: 103–105. ARMBRUSTER, W. S. 1993. Within-habitats heterogeneity in baiting samples of male euglossine bees: Possible causes and implications. Biotropica 25: 122–128. BAWA, K. S. 1990. Plant pollinator interactions in tropical forest. Annu. Rev. Ecol. Syst. 21: 399–422. BONILLA, M. A., AND G. NATES. 1992. Abejas euglosinas de Colombia. I. Claves ilustradas. Caldasia 17: 149–172. BROWER, J. E., J. H. ZAR, AND C. N. VON ENDE. 1997. Field and laboratory methods for general ecology, 4th edition. McGraw–Hill, Boston, Massachusetts. CANE, J. H. 2001. Habitat fragmentation and native bees: A premature verdict? Conserv. Ecol. http:// www.consecol.org/Journal/vol5/iss1/art3/ CLARK, D. A. 1992. Plant demography. In L. McDade, K. S. Bawa, H. A. Hespenheide, and G. S. Hartshorn (Eds.). La Selva: Ecology and natural history of a Neotropical rain forest, pp. 90–105. University of Chicago Press, Chicago, Illinois. DRESSLER, R. L. 1978a. New species of Euglossa from Mexico and Central America. Rev. Biol. Trop. 26: 167–185. ———. 1978b. An infrageneric classification of Euglossa with notes on some features of special taxonomic importance. Rev. Biol. Trop. 26: 187–198. ———. 1979. Eulaema bombiformis, E. meriana, and Mu¨llerian mimicry in related species. Biotropica 11: 144–151. ———. 1982a. Biology of orchid bees (Euglossini). Annu. Rev. Ecol. Syst. 13: 373–394. ———. 1982b. New species of Euglossa II. Rev. Biol. Trop. 30: 121–129. ———, AND R. OSPINA. 1997. Una nueva especie de Eulaema (Hymenoptera: Apidae) del Choco Colombia. Caldasia 19: 95–100. FABER-LANGENDOEN, D., AND A. H. GENTRY. 1991. The structure and diversity of rain forest at Bajo Calima, Choco region western Colombia. Biotropica 23: 2–11. FERNA´NDEZ, F. 1995. La diversidad de hymenopteros en Colombia. In J. O. Rangel (Ed.). Colombia diversidad bio´tica I, pp. 373–442. Universidad Nacional de Colombia,Bogota´, Colombia. GILBERT, L. E. 1980. Food-webs and the conservation of Neotropical diversity. In M. E. Soule´ and B. A. Wilcox (Eds.). Conservation biology, pp. 11–33. Sinauer Associates, Sunderland, Massachussetts. HERRERA, L. 1989. Colombia prehispa´nica VII: Costa del Oce´ano Pacı´fico y Vertiente Oeste de la Cordillera Occidental. http://www.banrep.gov.co/blaavirtual/letra-p/prehisp/cp13.html IGAC (INSTITUTO GEOGRA´FICO ‘‘AGUST´IN CODAZZI’’). 1977. Zonas de vida o formaciones vegetales de Colombia, vol. 13, no. 11, Bogota´, Colombia. JANZEN, D. H. 1971. Euglossine bees as long-distance pollinators of tropical plants. Science 171: 203–205. ———. 1974. The deflowering of Central America. Nat. Hist. 83: 48–53. ———. 1981. Bee arrival at two Costa Rican female Catasetum orchid inflorescences, and a hypothesis on euglossine population structure. Oikos 36: 177–183. ———, P. J. DEVRIES, M. L. HIGGINS, AND L. S. KIMSEY. 1982. Seasonal and site variation in Costa Rican euglossine bees in lowland deciduous and evergreen forest. Ecology 63: 66–74. KIMSEY, L. S. 1982. Systematics of bees of the genus Eufriesea. Univ. Calif. Publ. Entomol. 95: 1–123. ———, AND R. L. DRESSLER. 1986. Synonimic species list of Euglossini. Pan-Pac. Entomol. 62: 229–236.

Euglossines across an Intervention Gradient

529

KRESS, W. J., AND J. H. BEACH. 1992. Flowering plants reproductive systems. In L. McDade, K. S. Bawa, H. A. Hespenheide, and G. S. Hartshorn (Eds.). La Selva: Ecology, and natural history of a Neotropical rain forest, pp. 161–182. University of Chicago Press, Chicago, Illinois. MAY, M. L., AND T. M. CASEY. 1983. Thermoregulation and heat exchange in euglossine bees. Physiol. Zool. 56: 541–551. MORATO, E. F. 1994. Abundancia e riqueza de machos de Euglossini (Hymenoptera: Apidae) em mata de terra firme e areas de derrubada, nas vizinhancas de Manaus (Brasil). Bol. Mus. Emilio Goeldi, ser. Zool. 10: 95–105. OSPINA, R. 1998. Revisio´n de la morfologı´a genital masculina de Eulaema (Hymenoptera: Apidae). Rev. Biol. Trop. 46: 749–762. ———, AND J. C. SANDINO. 1997. Eulaema chocoana, nueva especie de abeja euglossina de la costa pacı´fica colombiana. Caldasia 19: 165–174. OTERO, J. T., AND A. SALLENAVE. Vertical stratification of euglossine bees (Hymenoptera: Apidae) in an Amazonian forest. Pan-Pac. Entomol. In press. OYAMA, K. 1993. Conservation biology of tropical trees: Demographic and genetic considerations. Environ. Update 1: 17–32. PEARSON, D. L., AND R. L. DRESSLER. 1985. Two-year study of male orchid bee attraction to chemical baits in lowland eastern Peru. J. Trop. Ecol. 1: 37–54. PERUQUETTI, R. C., L. A. O. CAMPOS, C. D. P. COELHO, C. V. M. ABRANTES, L. C. O. LISBOA. 1999. Abelhas euglossini (Apidae) de a´reas de Mata Atlaˆntica: Abundancia, riqueza e aspectos biolo´gicos. Rev. Bras. Zool. 16: 101– 118. POWELL, A. H., AND G. V. N. POWELL. 1987. Population dynamics of male euglossine bees in Amazonian forest fragments. Biotropica 19: 176–179. RINCO´N, M., D. W. ROUBIK, B. FINEGAN, D. DELGADO, AND N. ZAMORA. 1999. Understory bees and floral resources in logged and silviculturally treated Costa Rican rainforest plots. J. Kans. Entomol. Soc. 72: 379–393. ROUBIK, D. W. 1989. Ecology and natural history of tropical bees. Cambridge University Press, Cambridge, England. ———. 1992. Loose niches in tropical communities: Why are there so few bees and so many trees? In M. D. Hunter, T. Ohgushi, and P. W. Price (Eds.). Effects of resource distribution on animal–plant interactions, pp. 327– 354. Academic Press, San Diego, California. ———. 1993. Tropical pollinators in the canopy and understory: Field data and theory for stratum ‘‘preferences’’. J. Insect Behav. 6: 659–673. ———. 2001. Ups and downs in pollinator populations: When is there a decline? Conserv. Ecol. http:// www.consecol.org/Journal/vol5/iss1/art2 ———, AND J. D. ACKERMAN. 1987. Long-term ecology of euglossine orchid bees. Oecologia 73: 321–333. SANDINO, J. C. 1995a. Primer inventario de abejas euglosinas en la vertiente del Pacifico de suroccidente de Colombia: Diversidad, distribucio´n altitudinal y efectos de la perturbacio´n humana. Tesis de pregrado. Universidad del Valle, Cali, Colombia. ———. 1995b. Efectos agrı´colas sobre la distribucio´n de la diversidad de machos euglosinos en un gradiente de intervencio´n, Buenaventura, Valle del Cauca. Informe final, Biopacifico, Santafe de Bogota D.C. Colombia. ———. Are there any agricultural effects on the capture rates of male euglossine bees (Apidae, Euglossinae)? Rev. Biol. Trop. In press. SAS. 1989. SAS/STAT user’s guide, version 6, 4th edition. SAS Institute, Cary, North Carolina. TONHASCA, A., J. L. BLACKMER, AND G. S. ALBUQUERQUE. 2002. Abundance and diversity of Euglossinae bees in fragmented landscape of the Brazilian Atlantic Forest. Biotropica 34: 416–422. WHITTEN, W. M., A. M. YOUNG, AND D. L. STERN. 1993. Non-floral sources of chemicals that attract male euglossine bees. J. Chem. Ecol. 19: 3017–3027. WILLIAMS, N. H. 1982. The biology of orchids and euglossine bees. In J. Arditti (Ed.). Orchid biology II, pp. 121– 169. Cornell University Press, Ithaca, New York. ZAR, J. H. 1999. Biostatistical analysis, 4th edition. Prentice Hall, Upper Saddle River, New Jersey. ZUCCHI, R., S. F. SAKAGAMI, AND J. M. F. CAMARGO. 1969. Biological observations on a Neotropical parasocial bee, Eulaema nigrita, with a review on the biology of Euglossine. A comparative study. J. Fac. Sci. Hokkaido. Univ. VI Zool. 17: 271–380.