Skip to main content

Ten-a-day: bumblebee pollen loads reveal high consistency in foraging breadth among species, sites, and seasons

Cite this dataset

Timberlake, Thomas et al. (2024). Ten-a-day: bumblebee pollen loads reveal high consistency in foraging breadth among species, sites, and seasons [Dataset]. Dryad.


Pollen and nectar are crucial resources for bees, but vary greatly amongst plant species in their quantity, nutritional quality, and timing of availability. This makes it challenging to identify an appropriate range of plants to meet the nutritional needs of pollinators through the year, though this information is important in the design of pollinator conservation schemes.

Using DNA metabarcoding of pollen loads, we record the floral resource use of UK farmland bumblebees at different stages of their colony lifecycle, and compare this with null models of ‘expected’ resource use based on landscape-scale resource availability (pollen and nectar), to identify foraging priorities and preferences. We use this approach to ask three main questions: i) what is the foraging breadth of individual bumblebees?; ii) do bumblebees utilise a greater or lesser diversity of plant species than expected if they foraged in proportion to resource availability?; iii) which plant species do bumblebees preferentially utilise?

Individual bumblebees foraged from a highly consistent number of different plant taxa (mean: 10 ±0.37 SE per bee), regardless of their species, sampling site, or time of year. This high consistency in foraging breadth, despite large changes in the quantity, identity, and diversity of resource availability, implies a strong behavioural tendency towards a fixed range of foraging resources. This effect was most striking in April when foraging diversity was maintained despite very low landscape-level resource diversity.

Bumblebees used some plant taxa significantly more than predicted from their landscape-level floral abundance, nectar, or pollen supply, implying certain desirable characteristics beyond the mere quantity of resource. These included Allium spp. and Vicia spp. in April; Trifolium repens and Lotus corniculatus in July; and Cynareae spp. (thistles) and Taraxacum officinale in September.

Our results strongly indicate that resource quantity is not the only factor driving bumblebee foraging patterns, and that resource diversity and quality are also important factors. Thus, in addition to providing large quantities of floral resources, we recommend that pollinator conservation schemes also focus on providing a sufficient diversity of preferred floral resources, enabling pollinators to self-select a diverse and nutritious diet.

README: Ten-a-day: bumblebee pollen loads reveal high consistency in foraging breadth among species, sites and seasons

This dataset provides information on the phenology of floral abundance on three farms in Somerset, UK, as well as DNA metabarcoding data (relative read abundances/ RRA) for all of the plant taxa recorded in bumblebee's pollen loads on these same three farms. We use these datasets in combination to compare the availability of floral resources in the landscape (floral abundance data), with the actual use of these resources by farmland bumblebees (DNA metabarcoding data).

Description of the data and file structure

Dataset 1: Floral resource phenology data


This dataset provides a list of all plant species recorded flowering on the three study farms (ET, B & EM) and their estimated floral abundance on each day of the flowering season, expressed as the number of floral units per kilometer squared of farmland (all habitats included). The dataset is structured as a matrix, with plant species as rownames and days of the year as columns. Also included in the dataset are the values of nectar (µgrams of sugar /day) and pollen (µm3 volume) produced by an individual floral unit of each species. This enables the user to calculate nectar or pollen availability for any species on any day of the year by multiplying the values of nectar or pollen per floral unit by the floral abundance values of the plant species. Nectar and pollen values are from Baude et al. (2016) and Wright et al. (2024), respectively, whilst estimates of floral abundance on each farm are derived from Timberlake et al. (2019) who modelled the floral abundance of each species using a Generalised Additive Model based on data from transect surveys (see the original paper for further details).

Dataset 2: Pollen DNA metabarcoding data


This dataset provides a list of all bumblebee specimens captured during the study, specifying the farm they were captured on, the month in which they were captured (April, July or September), and their species identity. The dataset is structured as a matrix with individual bumblebee specimens as rows, and the columns showing all of the plant taxa that were recorded in the pollen loads of each bumblebee specimen. The values in the matrix show the relative read abundance (RRA) of each plant taxon which is a measure of the proportion of total sequence reads from each specimen that were identified to a given plant taxon. See the methods in the paper for more information. Information on each specimen (e.g. it's date and location of capture, the plant it was captured on etc) is listed in a separate sheet within the same workbook, alongside a metadata sheet to explain each variable in the dataset.

Sharing/Access information

This dataset is publicly available and the modified source data (derived from these original datasets) can be accessed via Github, along with a fully-reproducible analysis workflow.

Link to the Github source data and code:


Code and source data for a fully reproducible analysis workflow is available from the following Github repository:


The data consists of floral resource phenology values (floral abundance, pollen and nectar) collected on three farms during 2017. In addition, there is pollen DNA metabarcoding data showing the relative read abundances of plant taxa recorded on 187 individual bumblebees. Full details on the collection methodology are provided in the paper.


Natural Environment Research Council, Award: NE/L002434/1