Bees often focus their foraging effort on a few or even a single flower species, even if other equally rewarding flower species are present. Although this phenomenon – called flower constancy – has been widely documented during single foraging trips, it is largely unknown if the behavior persists over longer time periods, especially under field conditions with large spatial and temporal variation of resources. We investigated the pollen diet of individuals from nine different Bombus terrestris colonies up to six weeks, placed at either 0, 300 or 1000 m from nearest autumn sown oilseed rape. Adjacency to oilseed rape affected constancy, with a higher flower constancy in colonies far from fields. Overall, 23 % of the pollen foraging trips were flower constant, and the fraction of constant pollen loads did not change over time, although repeatedly sampled individuals that were flower constant once often showed different preferences at other sampling occasions, and the similarity between pollen loads collected by the same individuals dropped over time. This suggests that flower constancy and other learnt foraging strategies likely change in response to shifting floral resources. How rapidly it changes in response to changing resources, and if this differs between species as well as within species depending on e.g. size, should be the focus of future research. The average diversity of pollen from single foraging trips was around 3 pollen types, and we found indications of resource partitioning within the colonies resulting in higher colony-level pollen diversity. Keywords: flower fidelity, Bombus, foraging, seasonal shift, foraging preferences, Brassica napus.

The data was collected in Southern Sweden during spring/early summer 2017. Commercial bumblebee colonies (Bombus terrestris) were placed in three different areas, and in each of these areas in three different sites that were located next to, 300m from, or 1000m from nearest oilseed rape field. The study design is a subset of the one described in Yourstone et al. 2021, https://doi.org/10.1016/j.biocon.2021.109249. 

Individual bumblebee workers were marked and sampled for pollen in their pollen basket (if present). Each row in the dataset "Pollen_id_Yourstone_et_al_2023.csv" represents one such sample. At the first sampling occasion, bumblebees returning to the nests were caught and marked by gluing a number plate onto their thorax, and the pollen from their corbicula was collected if present. At subsequent sampling occasions both already marked bumblebees and unmarked individuals were captured, marked if needed, and sampled for pollen if present. The entrance tube was cleaned between sampling session but not between bumblebee individuals. Later in the season the work was focused on recapturing marked bumblebees, however, new bumblebees were also marked if pollen could be collected from them. Each monitored colony was visited at least once a week, unless it rained, or wind speed was > 6 m/s. Each monitoring session lasted for about one hour, and all three colonies within the same area were visited the same day. The time of sampling was between 9:00 and 18:00. The colonies were visited between 8th of May and 27th of June, but at the end of this period very few or no bumblebees could be sampled because the colonies were at the end of their cycles.

To know what plant taxa the pollen in the samples belonged to, each sample was mounted on a microscopic slide and stained with fuchsin. The microscopic slides were scanned and analysed with a pre-trained deep learning convolutional neural network, that classified each pollen grain in the samples, giving it a probability of identity of different pollen types. These probabilities were summed for each sample, and the resulting frequencies per pollen type were adjusted according to Olsson et al. (2021) https://doi.org/10.1111/2041-210X.13575 to reduce the effects of contamination and uncertain classifications. The description of pollen types and what plant taxa these represent can be found in Appendix C to the article this data is connected to.

There is also a code file that uses the data described above and produces Jaccard distances between all pollen samples. 

The data file Pollen_id_Yourstone_et_al_2023.csv can be opened with various softwares, including R and Microsoft Excel. 

The code file can be opened as a plain text file, but needs to be opened with the open source software R to be run.