Switching plant species visited by pollinators (partner flexibility), has been proposed as a behavioural mechanism able to attenuate the negative impacts of shifts in plant communities on pollination. However, it is unclear if the magnitude of such response is generalizable or depends on the environmental context. Moreover, the ability of pollinators to exploit plants with dissimilar traits (trait flexibility) has been overlooked, even though it can affect the spectrum of new partners available.

To shed some light on this problem, we quantified partner and trait flexibility in five communities from four different environments, from Alpine to semi-arid. We evaluated if the rate at which pollinators incorporated new plant species throughout the flowering season was similar across communities or context-dependent. Then, we assessed if pollinators changed the type of flowers visited and if such trait flexibility was related to their capacity to visit new plant species. Finally, we developed an agent-based model to explore if diet flexibility can protect pollination when the plant community changes. To this end, we used scenarios of phenological decoupling.

In general, pollinators switched interaction partners to cope with the temporal replacement of plant species. Yet, the magnitude of such behaviour varied across communities, probably in response to differences in the number of floral resources available. Also, pollinators were able to visit plant species with dissimilar traits, though both components of diet flexibility (partner identity and traits) did not necessarily covary. Thus, to have a full picture of pollinators’ diet flexibility we need to consider the floral traits of partners. Finally, our theoretical model shows that diet flexibility can protect pollination after shifts in plant communities; but that such positive effects are limited by trait-matching between co-occurring species.

Overall, our study highlights the importance of incorporating plant traits when evaluating the ability of pollinators to find new interaction partners. Besides, our simulation results suggest that diet flexibility may not unequivocally protect pollination against changes in plant communities, especially if they entail shifts in the characteristics of the floral assemblage, and hence, the ability of pollinators to find new interaction partners can be compromised.

Some sets of data from field work is already available in online resources. When this is the case the specific link or identification (e.g. dryad file, figshare) is provided.

The dataset contains two parts. (1) Observational data from three regions- Montane temperate alpine, Mediterranean alpine and Monte desert and (2) a folder with scripts and input data from an Agent based model for pollination. Datasets from field work contain information of plant-pollinator interactions on weekly basis, relative abundance of flowers and plant traits. When the source of data has been published elsewhere the specific link is provided. In the case of the agent base model, we include the script, functions and input data so that it can be run.