Protecting biodiversity and predicting how vegetation and animal communities respond to negative impacts of human activities (e.g., landscape use and soil degradation) are key scientific topics in this century (Cardinale et al., 2012; Foley, 2005). Scientists must understand the patterns and processes in biological systems and develop assessment and evaluation procedures that assure the maintenance of biological resources (Yoccoz et al., 2001). That assessment must include direct biological monitoring (Lutter et al., 2018), which can provide information about species identity (where the data is absent or scarce), ecosystem process, and habitat conservation status. With science-budget cuts threatening even more conservation projects and actions (Kowaltowski, 2021; Malakoff, 2020), monitoring actions may compete for funds with many other activities related to biodiversity protection. Therefore, optimizing research resources is mandatory to increase spatial and temporal coverage of biodiversity monitoring.

Biodiversity monitoring must be sufficient to reflect fluctuations and trends in natural populations to increase our ability to inform conservation priorities (Jetz et al., 2019). This, however, is not simple or trivial as monitoring programs are usually designed to balance two main key points: the sampling area (extension) and resources (human and financial). The first point (extension) may reflect when sampling is enough, i.e., our capacity to sample populations or communities in the area, but often the decision about the sampling extension is related to financial or capacity building (i.e., the human capacity available to process the biological data) (Magnusson et al., 2005). This is particularly relevant for invertebrates in tropical regions since biodiversity studies are biased towards vertebrates and temperate regions (Titley et al., 2017).

Euglossine bees, also known as orchid bees, are ecologically important forest-dependent pollinators and thus significant components among pollinators of Neotropical forests. Orchid-bee conservation is considered vital for natural and semi-natural areas (Aguiar and Gaglianone, 2012; Brosi, 2009; Brosi et al., 2007). Since the discovery of aromatic-scent components attracting orchid bees in the 1960s (Dodson et al., 1969), this group is frequently used to assess environmental impacts. Orchid bees respond to stressful factors, such as forest fragmentation, and are sometimes considered good bioindicators of environmental quality (Aguiar et al., 2014; Aguiar and Gaglianone, 2012; Brosi, 2009; Carneiro et al., 2021).

The Amazon forest is reported to have the richest fauna and the highest levels of endemism of orchid bees (Nemésio and Silveira, 2007). Previous studies with this group indicate that some biotic, such as vegetation types and plant phenology (Aguiar et al., 2014; Zimmerman et al., 1989) or general abiotic variables, such as precipitation, altitude, temperature, or humidity (Zimmerman et al., 1989; Abrahamczyk et al., 2011; Andrade-Silva et al., 2012; Aguiar and Gaglianone, 2012; Aguiar et al., 2014; Giangarelli et al., 2015; Mateus et al., 2015), can be related to orchid-bee communities. However, relationships among variables and stressors can change over time, environment and local conditions (Belsky and Joshi, 2019; Polatto et al., 2014).

The Reserva Florestal Adolpho Ducke (Ducke Reserve), maintained by the Instituto Nacional de Pesquisas da Amazônia on the outskirts of Manaus, Amazonas, has been the focus of studies of diverse organisms (plants and animals) in recent decades (e.g., Castellani and Freitas, 1992; Hopkins, 2005; Lima and Magnusson, 2006). In general, the biodiversity of Reserva Ducke is tightly linked to its soil being acidic and poor in nutrients, such as phosphorus, calcium, and potassium (Chauvel et al., 1987). Thus previous studies dealing with population or community spatial structure included soil variables to investigate the spatial distribution of organisms in the Ducke Reserve (Aguiar et al., 2006; Costa et al., 2005; Guedes et al., 2021; Magnusson et al., 2005). Soil variables, such as phosphorus and the sum of bases, can influence nectar chemical composition and plant volatile compounds. However, although those variables have been investigated for other bee groups, such as Bombus and Apis, and plant communities (Nunes et al., 2015; Ceulemans et al., 2017), such aspects have not been used to assess the composition of orchid-bee communities. Understanding the landscape-scale distribution of bees can provide a more realistic scenario for conservation actions (Boscolo et al., 2017; Kremen and Ostfeld, 2005; Vickruck et al., 2021).

Understanding population patterns is important, but sampling must not cause depletion of individuals sufficient to threaten species. This is important for organisms already suffering declines due to anthropogenic disturbance, such as bees (Cane and Tepedino, 2001). Determining the sample size sufficient to investigate the relationships between local orchid-bee assemblages and abiotic factors associated with species distribution in the landscape is crucial for species conservation and optimizing monitoring programs. In this study, we investigated whether (i) the sampling area (plot, line or grid) affects diversity indices (richness, abundance and composition) of orchid-bee assemblages to determine the spatial distribution of sampling that maximizes the cost-benefit ratio of collection effort in surveys of orchid bees. We also (ii) investigated how orchid bees are related to environmental predictors (trees and palms species composition, soil phosphorus, and the sum of bases) at a mesoscale (25 km2) in central Amazonia. We predicted that orchid bees would be more related to plant composition variation (direct factor), than soil nutrients (indirect factors).

Using the nested hierarchical design, we collected 828 individuals, which represents 84% of the individuals collected before the implementation of the Biodiversity Research Program (PPBio) site and deposited in the INPA Entomological collection (Table 1). The overall number of species sampled (34) was also high, representing 79% of the total number of species previously sampled and deposited in the INPA Entomological collection. Over the entire grid, we sampled a high number of individuals (401), but the number of species was similar on the grid, line, and plot (24, 20 and 20, respectively). However, some species were collected exclusively on the grid, line or plot (Table 1).

The shape of the species accumulation curves varied among the sampling areas. While the plot and line curves were steeply inclined, the species accumulation curve of the grid tended to asymptote (Fig. 2A). Despite the steeper species accumulation in the grid at the beginning (i.e., more species were sampled for the same number of individuals), the number of species sampled and expected between the three scales at maximum sampling effort showed high overlap (Fig. 2A). However, the sampling completeness was much better for the grid (0.96) than for line (0.87) and plot (0.86). Even after controlling the variation between the number of individuals sampled, the grid on average, sampled 10% more species than the other sampling schemes (Fig. 3B).

Fig. 2.

(A) Individual-based species accumulation curves and (B) Sample coverage for the three sampling scales, grid, line, and plot.

(0.18MB).

Fig. 3.

Mean number of species and individuals (triangle) of bootstrap samples for the three sampling scales. The whisks represent 95% confidential intervals.

(0.14MB).

The pattern was different at the sampling unit scale. The number of species and individuals per sampling unit on grid scale was higher than line and plot, which no difference among those last (Fig. 3). Interestingly, the variation in sampling grain did not affect the number of species and individuals sampled (plot and line comparison). In contrast, the sampling scale was important, with more individuals and species sampled at the grid scale.

Each sampling scheme (grid, line and plot) captured a different assemblage species composition, but the sampling heterogeneity was very similar between sampling areas (Fig. 4, Table 2). There is no clear evidence for male depletion, even for the plot scale. The number of sampled individuals varied between each sampling day, even for the grid, which was sampled at different plots every day (Fig. 5).

Fig. 4.

NMDS 2D solution-based Bray–Curtis distance of Euglossine bees sampled at different spatial scales. Each point represents a similar sampling effort installed per plot/day.

(0.24MB).

Fig. 5.

Mean and 95% confidence intervals for each sampling day and sampling scale. All samples were undertaken on six consecutive days.

(0.15MB).

Trees and palms species composition, estimated here as the NMDS ordination axis (Wald-value = 4.151; p = 0.544), and sum of bases (Wald-value = 4.438; p = 0.472) were not related with orchid bees composition. However, orchid bees species composition was related with soil phosphorus content (Wald-value = 7.088; p = 0.002). Most bee species were sampled along the phosphorus gradient, except for Eufrisea ornata, Euglossa parvula and Eulaema polyzona, which seem to be restricted to areas richer in phosphorus (Fig. 6). However, except for E. parvula, the other two species were sampled in only one plot and their association with areas with higher phosphorus content should be considered cautiously.

Fig. 6.

Distribution of Euglossine species along Phosphorus content gradient in study plots at Reserva Florestal Adolpho Ducke, Manaus, Brazil.

(0.25MB).

Scent traps can be highly effective in collecting orchid bees, but the sample design must be as effective as possible without depleting organisms. As we observed in our study, we captured ∼80% of the orchid bee species deposited at the INPA Entomological collection using the grid sampling design, yet did not evidence a number of individuals captured during our sampling campaign. We noticed different assemblage compositions but similar composition heterogeneity between sample scheme. At mesoscale, orchid bee species composition was only related to soil phosphorus content, showing no pattern associated with palm and trees species composition. Those findings, especially the last, might shed some light on indirect effects of environmental variables on orchid bees species composition.

Euglossine bees are undoubtedly a relevant group for evaluating landscape effects, especially those related to monitoring conservation efforts or evaluating anthropic disturbances (Aguiar et al., 2014; Aguiar and Gaglianone, 2012; Brosi et al., 2007; Carneiro et al., 2021). The possibility to collect those male bees by standardized and efficient sampling made possible a substantial amount of data related to their distribution, diversity, and abundance. Very few works analyzed the sufficiency of a sample scale and the relation of orchid bees’ composition with the environment. Our results with a standardized protocol reveal that although the number of estimated species richness overlaps, the sampling coverage was higher for the grid compared with the other sampling schemes. The grid design was also more effective ignoring the species identity, suggesting that distributing the traps in larger areas collects more species and individuals per trap than concentrating the same number of traps on trails or plots. Thus, collecting along the grid was more efficient than focusing the collection effort on a trail or plot.

The males of orchid bees have a high detection ability being highly attracted to aromatic scents. Although it is not entirely clear the maximum distances at which those bees can detect bait essences, evidence was found that they can achieve flight distances from up to 50 km (Pokorny et al., 2015), yet, their ability to cross open areas is limited requiring conservated landscape areas instead of isolated patches (Powell and Powell, 1987). The high attractiveness of scent traps could lead us to think that leaving traps for many days or in places that do not favor the displacement of pollinators could decrease bee populations in an area. We did care about this data, yet fortunately, our results do not indicate this pattern. There is variation in the number of bees captured over the five consecutive days, but there is no clear decreasing trend. The number of sampled individuals was similar among sampling schemes, even for the grid, which was sampled at different plots every day. It is possible that the high dispersion ability acted to minimize the depletion effect. The high dispersion capacity of some species, associated with the high detection power, may maintain the number of individuals sampled daily. Thus, this seems to be an efficient method that may have limited effects on the population level during short periods (five days) of continuous sampling.

We also evidenced that correlations between bees and environmental variables are not always obvious and can be site-specific. Previous works correlated the orchid bee assemblages with different variables such as flowering plants or local physiognomies (Ackerman and Roubik, 2012; Aguiar and Gaglianone, 2012; dos Santos et al., 2020), temperature, altitude, precipitation, and elevation (Aguiar and Gaglianone, 2012; Armbruster and Berg, 1994; Armbruster and McCormick, 1990; dos Santos et al., 2020). Our study did not find an association between bee assemblage composition with trees and palm species composition. Instead, we found an association with the soil phosphorus content. At the Reserva Ducke, soil variables shape the local-scale distribution of many animals and plants communities and thus is a promissory site-specific variable (Aguiar et al., 2006; Costa et al., 2005; Guedes et al., 2021; Magnusson et al., 2005).

Other nutrients, including phosphorus, have already been investigated for other bee groups such as Bombus and Apis (Ceulemans et al., 2017; Nunes et al., 2015). Ceulemans et al. (2017) found that artificial nutrient enrichment was associated with altered nectar and pollen chemical composition, increasing larval mortality of Bombus. As far as we know, the relationship between soil phosphorus content and orchid bee composition was never related before. As phosphorus was the most relevant predictor, future works should try to understand the possible causality with the orchid bees and the mechanisms between them.

In our study area, phosphorus evidenced differences in some species’ occurrence patterns. E. ornata, E. polyzona, and E. parvula are restricted to richer phosphorus areas. However, the first two were sampled in one plot, and all three species had fewer sampled individuals. It is common for crop plants to use phosphorus application to increase productivity; however, its effects are far beyond soil and plant; bee visits are augmented by this relationship as nectar can be more attractant (Karunakaran et al., 2021; Vaudo et al., 2022). In native plants, we could expect the same relationship due to this high bee dependence on vegetation due to the use of resources (nectar, pollen) or microclimate conditions (temperature, humidity) (Polatto et al., 2014; Prado et al., 2021) the phosphorus variation may influence the foraging patterns of male orchid bees. Nevertheless, phosphorus availability is particularly low in central Amazonia (Quesada et al., 2010). Our results reinforce that slight changes in edaphic variation would result in disproportionately more significant indirect or direct effects on vegetation dynamics (Cunha et al., 2022).

The community structure of male orchid bees in the Neotropics follows the general pattern in which there are many individuals from a few species and a large number of species represented by only a few individuals (Janzen, 1971; Nemésio and Silveira, 2007; Sofia and Suzuki, 2004). Thus, optimizing the sampling effort is crucial for this group's preservation. It is also essential to consider that landscape can be perceived differently by the different species of orchid bees, as some species can act as bioindicators of environmental quality or degraded areas (Peruquetti et al., 1999). We still lack known orchid bees’ distribution in Central Amazon and how this group relates to landscape effects or environmental quality. Due to the growth of deforestation in the Amazon Forest, especially in the so-called “Arc of Deforestation,” these bees could provide quick and valuable information about landscape quality (Roubik and Hanson, 2004; Tonhasca et al., 2002). We thus need much more information regarding orchid bees to provide conservation efforts for this group. Our work highlighted the need to consider soil and nutrient variables other than vegetation and distribute scents traps in larger areas instead of in small plots.