Short-term lab assessments and microcolonies are insufficient for the risk assessment of insecticides for bees.

Risk assessment studies addressing effects of agrochemicals on bumblebees frequently use microcolonies. These are queenless colonies consisting of workers only in which typically one worker will lay unfertilized male-destined eggs. In the first tier of risk assessment for bees, short-term laboratory experiments (e.g. microcolonies) are used, the results of which will determine whether higher tier (semi-)field experiments are needed. To evaluate the suitability of microcolonies for risk assessment, a direct comparison between different assessment methods for the neonicotinoid pesticides acetamiprid and thiacloprid was made: microcolonies and queenright colonies under short-term laboratory conditions, queenright colonies under long-term laboratory conditions, and queenright colonies under field conditions. Here, we demonstrate that results from microcolonies contradict results from queenright colonies. While thiacloprid negatively impacted gyne production in queenright colonies, it had a positive effect on microcolony size. By contrast, thiacloprid had no significant effect on fitness parameters of queenright colonies under short-term laboratory conditions when mostly workers are produced. These results thus highlight both the need for long term assessments, allowing evaluation of gyne production, and the risk of reaching erroneous conclusions when using microcolonies. The negative effect of thiacloprid on colony fitness was confirmed under field conditions, where thiacloprid affected the production of reproductives, colony weight gain, worker weight, and foraging behaviour. For acetamiprid, a negative trend on colony fitness could only be shown in a field setup. Therefore, field-realistic setups, which allow colonies to forage freely, are most appropriate to assess sublethal effects of pesticides affecting behaviour and learning.