The decision-making process of leafcutting bees when selecting patches

Abstract

Change in land configuration is an important driver of pollinator decline. Understanding patch selection by bees in fragmented landscapes has therefore become imperative to guide the design of habitats that support pollinators and ensure their conservation. This is especially true for solitary bees that make up most bee species in the world. To elucidate the decision-making process of a solitary bee when selecting patches, we tested four models of patch attractiveness that differed in the role of patch size and isolation distance in the selection process. In these models, bees used both patch size and patch distance, only patch distance, or chose randomly among patches. When patch size was included, bees could estimate patch resources fully or partially. An experiment with a centre patch, surrounded by four peripheral patches of different sizes and distances from the centre, provided observed transition data to test against predictions derived from each of the models. The alfalfa leafcutting bee, Megachile rotundata, does not move randomly among patches. This bee uses both patch size and isolation distance when selecting a patch but can only evaluate patch resources partially. This knowledge can guide the design of habitats in fragmented landscapes to facilitate solitary bee conservation.

1. Introduction

Change in land configuration is one of the most important global drivers of pollinator decline [1,2]. Knowledge of patch selection in fragmented landscapes is crucial to guide the design of habitats that support pollinators and ensure their conservation. Understanding patch selection and patch utilization by bees has therefore become imperative [3]. Moreover, patch selection by wild and managed bees in crop fields will influence gene flow [4], and the spread of genetically engineered genes [5]. As solitary bees make up the majority of bee species in the world (over 85%) [6], and provide valuable pollination services for both native and agricultural ecosystems, there is an urgent need to understand how solitary bees select patches over the landscape.

Recently, four probability models of bee movement that differed in theories of patch attractiveness were developed to test patch selection by bees [7]. Patch size and/or patch distance define attractiveness in these models, and both are considered in models I and II (see details in the methods section). In model I, each point in the patch is assumed to emit an attractant, and the pollinator perceives the entire patch area. In model II, a pollinator views the patch edge-on and only perceives its diameter. In model III, attractiveness only depends on distance, and this model is the closest approximation to the nearest neighbour model. In model IV, all patches are equally attractive, illustrating a random choice model. Understanding which model best describes patch selection, and potential variation among bee species in their selection process, would improve the design of landscapes for bee conservation.

In a previous experiment, Bombus impatiens L. was shown to use both patch size and distance when selecting patches (model I) and could estimate the total amount of resources available in a patch [7]. However, many differences exist in the foraging behaviours and cognitive abilities of social and solitary bees [8–11]. For example, when comparing social (B. impatiens) and solitary (Megachile rotundata F.) bees foraging on Medicago sativa, Megachile rotundata did not exhibit directionality of movement, travelled shorter distances, and set more seeds per visit than B. impatiens [8,12]. Such differences between a social and a solitary bee could extend to the patch selection process.

Here, we examined the patch selection process of a solitary bee, M. rotundata, foraging in patches of Medicago sativa L. This cavity nesting bee nests in aggregation which facilitates its use as a managed pollinator [13]. Because it uses pieces of leaves to build a cell, each with an egg and provisions, it is commonly called a leafcutting bee. We recorded bee transitions from a centre patch to each of four peripheral patches of two sizes and located at two distances from the centre. We identified the best patch selection model by comparing observed against predicted transitions under each of the four models. Megachile rotundata uses spatial information over patchily distributed resources when selecting a patch. We discuss how such knowledge can guide the design of habitats in fragmented landscapes to facilitate bee conservation.

2. Material and methods

(a) . Experimental design

The study took place at the University of Wisconsin Walnut Street Greenhouse in Madison, Wisconsin, USA. Each experiment consisted of four peripheral patches of alfalfa of two sizes and located at two distances from a centre patch. The centre patch and two peripheral patches measured 91.4 × 91.4 cm and contained nine potted plants (small), while the other two peripheral patches were 137 × 137 cm and contained 20 potted plants (large). One of the small and one of the large patches were located 91.4 cm diagonally from the centre patch (near), while the other small and large patches were placed 183 cm away (far). This spatial configuration allowed bees to forage on a large near (LN), a small near (SN), a large far (LF) and a small far (SF) patch (figure 1). This set-up represents a scaled down version (to fit in a greenhouse room) of a similar field experiment that examined bumblebee transitions. Keeping similar scales facilitated comparisons between bee species.

Figure 1.

Experimental design. Four peripheral patches of two sizes located at two distances from a centre patch and set up in two configurations. The same size patches were set up (a) diagonally across the centre patch (opposite-side configuration) or (b) on the same side of the centre patch (same-side configuration). The design included a large near (LN), small near (SN), large far (LF) and small far (SF) patches.

At the beginning of an experiment, potted flowering plants were set up for each patch, and plants were replaced as needed to maintain a constant floral density among similar-size patches throughout an experiment. We conducted three separate experiments. Experiments 1 and 2 occurred in the spring and autumn of 2020, respectively, and had an opposite-side configuration, where peripheral patches of the same size were located opposite each other, diagonally (figure 1a). Experiment 3 took place in the spring of 2021 and was set up with a same-side configuration, where the two large patches and the two small patches were situated on the same side of the centre patch (figure 1b). The use of different configurations helped generalize the results of the study.

(b) . Bee observations

Commercially available cocoons were incubated in 1 l jars (half filled) in a growth chamber set at 30°C. Every other day, for the duration of the experiment, emerged bees from two jars were released in a collapsible aluminium cage (60 × 60 × 60 cm), set up in the greenhouse room, to allow mating. Prior observations, approximately 10–20 female bees (a limited number to facilitate observations) were moved to the greenhouse room. Four styrofoam nesting blocks (30 × 50 cm) were placed along the northern and western sides of the room to provide female bees with nesting cavities. Only transitions made by female bees were recorded. Observations typically took place between 09.00 and 16.00, when bees were active.

To facilitate the tracking of bees, fluorescent pigment powder of different colours was applied daily to the bees’ thorax with a paintbrush. Recently dusted bees were allowed to forage but were not immediately followed. Observers followed bees foraging in the centre patch and recorded a transition when a bee left the centre patch and moved to a peripheral patch and continued to forage in the new patch. For each transition observed, the date, time and patch type the bee moved to, were recorded. In experiments 1, 2 and 3, transition data were collected over 10, 13 and 11 days, respectively. The frequency of transitions to each of the four peripheral patches was calculated for each experiment.

(c) . Models of patch attractiveness

We have previously developed four different probability models of bee movement between patches that differ in theories of patch attractiveness [7]. The derivations of these models are presented in Fragoso et al. [7]. Discs of radius R are used as patches in the mathematical construction of the models for simplicity and tractability. In model I, the derivations show how attractiveness is proportional to the disc area (patch size) divided by squared distance (table 1). In model II, attractiveness is proportional to the diameter of the patch over distance (table 1). In model III attractiveness is inversely proportional to the square root of the distance between the centre of the patch and the bee (table 1). In model IV, all patches are equally attractive, the equivalent of a purely random patch choice model (table 1). Square patches were used in the experiments for ease of implementation, and thus, model predictions are an approximation to the true attractiveness of the patches in the experiment.

Table 1.

The four distinct models of patch attractiveness. Models differ in how patch size and patch distance affect attractiveness. A = patch area, d = distance between patches, R = patch radius.

model	factors determining attractiveness	formula
I	distance and patch area	A/d2
II	distance and patch diameter	2R/d
III	distance	1/d2
IV	random choice	1/4

(d) . Data analysis

For each of the four models, we calculated the predicted number of transitions to each of the peripheral patches by inserting the spatial dimensions of the experimental design into the respective model's equation. We then compared predicted and observed transitions, based on frequencies, using chi-square goodness of fit tests. For these tests, a model that represents a good fit to the data has a probability value > 0.05.

3. Results

(a) . Predicted transitions

Model I predicted more transitions to the large near patch, followed by small near, large far and small far patches (figure 2a). Compared to model I, model II predicted a lesser proportion of transitions to the near (both small and large) patches, and a greater proportion to the far (both small and large) patches (figure 2a). Model III anticipated the most transitions to the small near patch, while with model IV bees moved equally frequently among the four patches (figure 2a).

Figure 2.

Predicted and observed transition probabilities. The transition probabilities to each patch type (a) predicted under each of the four models and (b) observed in the opposite-side configuration (experiments 1 and 2 combined, N = 154), and the same-side configuration experiment (experiment 3, N = 101).

(b) . Observed transitions

We recorded a total of 255 transitions from the centre patch to the peripheral patches. Among these, 35 transitions and 119 transitions were observed, respectively, in the first and second experiments with opposite-side configuration, and 101 transitions were observed in the third experiment with same-side configuration. Transitions from the first and second experiments (opposite-side configuration) were combined prior to running the chi-square test, because the first experiment did not meet the chi-square assumptions due to its smaller sample size.

(c) . Model selection

For the opposite-side configuration (experiments 1 and 2 combined), model II best explained the observed transition data (table 2). In experiment 3, with the same-side configuration, only model IV was rejected (table 2). However, model II was a good fit to the data for both configurations, suggesting M. rotundata used both distance and patch size when selecting a patch, with partial resource estimation. The fewer number of models rejected with the same-side configuration implies that spatial configuration of the patches affects the decision-making process of the bees.

Table 2.

Model selection. Chi-square values and associated p-values for comparisons between observed and expected transition probabilities for each of the four models of patch attractiveness (described in the text). Experiments (Exp.) 1 and 2 were set up as opposite-side configuration, and experiment 3 as same-side configuration (see text for details). A model that is a good fit to the data has a probability value > 0.05, which are italicized.

exp.	config.	model I		model II		model III		model IV
		X2	p	X2	p	X2	p	X2	p
1 and 2	opposite	18.24	<0.001	6.48	0.090	18.84	<0.001	17.17	<0.001
3	same	6.59	0.086	6.54	0.088	4.51	0.211	18.80	0.000

4. Discussion

Megachile rotundata uses both isolation distance and patch size when selecting patches, confirming their ability to integrate two patch attributes in their decision-making process. Neither the social [7] nor the solitary bee species moved to the nearest neighbour patch (model III), or moved randomly among patches (model IV). Thus, despite their small brains, bees are capable of sophisticated learning and complex problem-solving [10,14,15]. Furthermore, these results indicate that both patch size and distance between patches should be considered when designing landscapes to improve bee conservation. The leafcutting bee preferred the near (both small and large) to the far patches, while the bumblebee strongly favoured the large near patch; large patches at close distances would therefore optimize bee foraging.

Bee species with larger brains are suggested to have greater cognitive abilities, and thus higher learning capacity, than bee species with smaller sized brains [9]. Bumblebees are larger than leafcutting bees, implying a greater brain size. Evidence on the cognitive ability of leafcutting includes their ability to discriminate among colours [9,16,17] and among geometric patterns when foraging [11] and selecting nesting sites [18]. They are also capable of solving new tasks in response to changes in their environment [19]. We have much to learn about leafcutting bee navigation, spatial memory and visual perception because most of these studies have been done using social bees [20–24]. Future research should examine whether differences in cognitive abilities between these two bee species can explain their differential estimation of resources available in a patch.

Patch configuration affects the decision-making ability of M. rotundata, as was previously observed for B. impatiens [7]. Here, only model II explained the transition data under the ‘opposite-side’ configuration, but other models, in addition to model II, could not be discarded under the ‘same-side’ configuration. At smaller scales, spatial configuration has also been shown to affect bee behaviour [20,24]. We are currently testing whether the pairwise presentation of patches, which differs between the two configurations, could help explain this difference. Adjacent pairs of patches vary in both patch size and patch distance only in the ‘opposite-side configuration’, and we are examining whether having both criteria vary simultaneously facilitates the decision-making process of the bees.

When designing landscapes for bee conservation, land managers must consider the ability of bees to use spatial information over patchily distributed resources. Both patch size and isolation distance affected the patch selection process of the bee species tested so far, although resources were estimated differently. A landscape consisting of large, nearby patches would be used most efficiently by these two bee species and could minimize gene flow. In addition, large patches attract more bees [21,22] which tends to increase plant reproductive success [12,23]. While this study used patches of a single-plant species and one bee species, future studies should examine patches of mixed-plant species and test other bee species. An efficient landscape for bee preservation must consider not only the types of plants offered, but also their spatial distribution, and the fact that bee species vary in how they use spatial information over patchily distributed resources.

Data accessibility

Data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.vdncjsxzg [25].

Authors' contributions

F.P.F.: conceptualization, formal analysis, investigation, visualization, writing—original draft and writing—review and editing; J.B.: conceptualization, funding acquisition, project administration, supervision, writing—original draft and writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This study was supported by a Biotechnology Risk Assessment Grant Program from the National Institute of Food and Agriculture (competitive grant no. 2018-33522-28707) and by funds from the United States Department of Agriculture-Agricultural Research Service to J.B. F.P.F. was supported in part by an appointment to the Agricultural Research Service (ARS) Research Participation Program administered by the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy (DOE) and the U.S. Department of Agriculture (USDA). ORISE is managed by ORAU under DOE contract number DE-SC0014664. All opinions expressed in this paper are the authors' and do not necessarily reflect the policies and views of USDA, ARS, DOE, or ORAU/ORISE.