Yesterday (26 March 2013), Defra published its field research on the effects of neonicotinoids on bumblebee colonies (Thompson et al 2013). Here is my understanding of the results.
The study finds fairly consistent exposure to neonicotinoids in free-living bumblebee colonies. One or more of the chemicals were found in at least some colonies at all three test sites, including the control site, which was next to a 6.5 ha field of oilseed rape not treated with neonicotinoid.
The actual neonicotinoids found in pollen and nectar collected by the colonies were not the same as the chemicals used on the treated fields the colonies were next to. Thiamethoxam was found at the highest levels, although it was not used on the treated fields included in the experiment. Use of thiamethoxam has rapidly increased in the UK in the last three years – 250,000 ha were treated with it in 2010.
This finding is important because it shows very clearly that this design of field trial, with bumblebee colonies placed at the edge of treated or untreated fields (fields were 6.5 to 12 ha in size) is not sufficient to compare the effects of neonicotinoid exposure against control ‘untreated’ colonies. I am not at all surprised by this. It is well known that bumblebees forage over several kilometres.
Levels in bumblebee-collected nectar and pollen ranged from detectable, but too low to measure quantities, up to an average of 2.4 μg/Kg (equivalent to parts per billion) of thiamethoxam in nectar for colonies at one site. Average levels of 0.7 μg/Kg of thiamethoxam were found in pollen at two of the three sites. This is lower than the 6 ppb used in pollen by the Whitehorn et al experiment, where they measured an 85% drop in queen production (Whitehorn et al. 2012), although Whitehorn used imidacloprid, which was only found below the levels of quantification at one of the sites in this study. Different neonicotinoids do not necessarily have the same toxic or sublethal effects.
There are not consistent and strong statistical correlations between the level of exposure to different neonicotinoids and colony mass or queen production, at these levels of exposure. This is why Defra says their research finds ‘low risk to bee populations’ from current use of neonicotinoids.
The study uses two statistical approaches. Some significant effects are found, but there seem to be no cases for which a strong statistical signal is found using both approaches. For example, one approach (potentially powerful but with strong assumptions) finds an effect of clothianidin in nectar on colony mass at the end of the study. The other approach (cruder but with fewer assumptions) finds effects of all measures of neonicotinoid exposure on colony mass after 3-4 weeks, but not at the end of the study. On queen production, little overall effect is found by either approach.
If the effect of neonicotinoids on bee colony performance was strong and clear, a range of different statistical approaches would detect it. The effect here isn’t massive, or overwhelming like the effect on queen production found by Whitehorn et al. Even at the site with the least productive colonies, they produced an average of 16 queens/ colony, which is comparable to buff-tailed bumblebee colonies measured in other situations. Control colonies in Whitehorn et al.’s study produced 13 queens on average, for example.
However looking at the data in the report (Figure 7, page 26) it seems there could be some subtle impacts on bumblebee colony performance at these exposure levels. I would really like access to the raw data.
The reporting of statistical results in this study is difficult to interpret. The main results table (Table 7) does not explain what the different rows mean. Results are reported as the percentage of models from 1000 partially simulated datasets that found ‘a significant effect’. Some linear model results are given in the Appendix (p 51 onwards), but they are not well explained. It is not clear to me how these relate to the results presented in Table 7. We can’t easily see the strength of interactions between different response variables in these different test runs. Does the effect of clothianidin in nectar change according to the exposure to thiamethoxam in pollen, for example? What happens if you don’t simulate data for the colonies exposed at levels below the quantifiable threshold?
The study monitored 60 buff-tailed bumblebee colonies Bombus terrestris, with 20 colonies at each of three sites in Lincolnshire or Yorkshire. It shows, as expected, that buff-tailed bumblebees are not feeding exclusively on oilseed rape even in intensive arable landscapes in the early summer. Only 13-26% of the pollen collected by the bumblebees was from oilseed rape. A similar proportion may also be true for nectar.
Overall, the consistency and levels of exposure to neonicotinoids in free-living bumblebee colonies revealed by this study are not surprising, given the use of these chemicals in the landscape. As a society, the results should make us think carefully about the impact of our agricultural practices on wild bees and other insects. The bee species used, the buff-tailed bumblebee, is common, widespread and not known to be declining. The findings should prompt urgent and very careful consideration of the effects of such consistent exposure on other, more vulnerable species.
Thompson, H., Harrington, P., Wilkins, W., Pietravalle, S., Sweet, D. and Jones, A. (2013) Effects of neonicotinoid seed treatments on bumble bee colonies under field conditions. Food and Environment Research Agency report. Available from: http://www.fera.defra.gov.uk/scienceResearch/scienceCapabilities/chemicalsEnvironment/documents/reportPS2371Mar13.pdf. Accessed 27 March 2013.
Whitehorn, P.R., O’Connor, S., Wackers, F.L., and Goulson, D. (2012) Neonicotinoid pesticide reduces bumble bee colony growth and queen production. Science 336, 351-352.