Research in focus: Country-specific effects of neonicotinoid pesticides on honey bees and wild bees. By Woodcock et al (2017), Science 365 (6345), pp. 1393-1395. DOI: 10.1126/science.aaa1190
On the 29th of June, the Centre for Ecology and Hydrology (CEH) published the largest field study to date examining the effects of neonicotinoid insecticides on bees. This much-anticipated study was funded by pesticide manufacturers Syngenta and Bayer, and carried out independently by the Centre for Ecology and Hydrology (CEH) after intense scrutiny of the methodology by researchers at the University of Cambridge.
As with much of the scientific literature on neonicotinoids, the results of this study have been debated fiercely with some controversial reporting from both sides. Bayer have since published a partial re-analysis doubting CEH’s assessments, while Syngenta claim the results to reflect positively on neonicotinoid usage and the mitigating impact of good beekeeping husbandry planting wildflowers around treated fields.
Here, Bumblebee Conservation Trust’s Head of Science, Dr Richard Comont, carefully explores the findings in detail and gives his take on what the results show…
This study sets out to establish the in-field effects of neonicotinoid-treated oilseed rape (Brassica napus, OSR) crops on bees in Europe (Germany, Hungary, and the UK). Bees studied were the Honey bee (Apis mellifera), Buff-tailed bumblebee (Bombus terrestris), and the solitary Red Mason bee (Osmia bicornis). Standard commercially-grown winter oilseed rape was seed-treated with either clothiniadin, thiamethoxam (both neonics) or nothing (control). Sites across the three countries were arranged in triplets, with one plot of each treatment in each triplet, to minimise the effects of site location on the treatment results. No other OSR crops were grown within 1.5km of the study sites, and any flowering crops nearby would not have been treated with neonics due to the ongoing EU-wide ban. Six Honey bee hives, 12 commercial bumblebee nests, and 50 Red Mason bee cocoons with two trap nests were placed at each site.
A number of colony measurements were taken during the year of exposure and in the following spring (number of queens produced, etc) and colony stores (pollen stores, nectar/honey stores, pollen loads from worker bees, and honey stomach contents) were examined for the presence of the pesticides. Pesticides expressed in the crop were also measured. These data were analysed to examine any effects of the pesticides across treatments, species and countries: in particular, modelling was carried out to test whether the presence of pesticides better explained colony measurements than any alternative explanations such as country-level effects.
The paper avoids the pitfalls of previous experiments by having field-realistic levels of exposure to neonics (it’s a field experiment with no artificial feeding of pesticides), and by being big enough to detect small-scale effects (33 field sites across three countries). It also does well by including bumblebees and solitary bees, rather than just honeybees, as honeybees have an unusual and resilient colony lifestyle and have been found to react differently to other bees in previous trials (Rundlöf et al. 2015; Henry et al. 2015).
As the experiment was a trial of the normal field effects of the pesticides, neonic seed treatments were standard, off-the-shelf mixes available to farmers (before the current ban). Each seed treatment thus also contained other crop protection products, principally fungicides, which are known to affect the effect of pesticides on bees. Therefore this study became more a comparison of the seed treatments Cruiser (thiamethoxam and the two fungicides fludioxonil & metalaxyl-M) and Modesto (clothianidin, the fungicides thriam & prochloraz, and the pyrethroid beta-cyfluthrin) to a control (the fungicides thriam & dimethomorph in Germany & Hungary or the fungicides thriam prochloraz in the UK), rather than a strict comparison of the neonicotinoid pesticides alone. All sites also received typical commercial inputs of fertilizer and pesticides (e.g. lambda-cyhalothrin), with inputs standardised across each triplet.
Honey bee beekeeping practices were kept as standard for each country, but there are differences between countries. Principally, the subspecies of Honey bee used, which was A. mellifera carnica in Germany and Hungary, and the Buckfast strain in the UK (a hybrid strain containing a mixture of A. m. ligustica, A. m. mellifera, A. m. anatoliaca, A. m. cecropia, A. m. monticola, & A. m. sahariensis), winter feeding (fondant or sugar solution), and Varroa mite treatments and levels of infection (Germany and Hungary use formic acid, UK oxalic acid, UK had significantly higher infection levels).
Additionally, German Honey bees were found to have a significantly lower proportion of OSR pollen in hive stores than did UK or Hungarian hives, and UK Honey bees foraged from a significantly smaller number of plant species than either Hungarian or German hives.
The headline findings were that pesticides were detrimental on reproduction for Buff-tail bumblebees and Red Mason bees, but that effects on Honey bees were mixed, with some negative effects but some apparently positive.
The quantity of neonic pesticide residue in the bee nests at each site was not correlated with the seed treatment at each site, and residues of imidocloprid, a neonic not used in the trials and banned on flowering crops under the existing restrictions, were discovered in bee colony stores at three sites. Additionally, thiamethoxam residue was found at one control and one clothianidin site and clothianidin was found at one control and two thiamethoxam sites (this may not represent contamination as clothianidin is a metabolite of thiamethoxam). This is thought to be linked to historical pesticide usage on the sites.
Clothianidin was significantly more likely to be expressed at a detectable level in plants which had had a clothianidin seed treatment, providing an exposure route for this pesticide, but this was not found for thiamethoxam.
The majority of lab-based studies to date have found serious sub-lethal effects on individual honey bees, but field trials have found no discernible significant effects on honey bee hive health (Pilling et al. 2013; Cutler et al. 2014; Rundlöf et al. 2015; Henry et al. 2015) suggesting that hives can withstand the effects even if some workers are adversely affected. This pattern largely holds true here.
The performance of Honey bee hives over the flowering period of the OSR crop was quantified as the peak number of workers, eggs, larvae, pupae, male brood, and storage cells produced while the crop was in flower. Overwintering performance was then quantified as overwintering hive survival, worker numbers, total brood (sum of eggs, larvae, pupae & male brood), and stored hive products. UK overwintering survival was too low for analysis so only German and Hungarian Honey bees were used for the overwintering survival analysis.
These data were then modelled against neonic treatments. Models always included country, the percentage cover of OSR around each site, percentage cover of arable land around each site, and the maximum neonic residue concentration (natural logs) in both colony stores and in the crop as covariates, as these were found to explain some of the variation between sites. The baseline null model (which assumes all variation is explained by these five factors alone) was then compared to a seed treatment model (the null model’s five explanatory variables plus treatment type), and an additive seed treatment model (null model plus an interaction, treatment*country), and the two seed treatment models were compared to each other. The aim of this step was to examine whether seed treatments had an effect, and whether this was a general effect or mediated by country-level effects.
During the flowering period, significant effects overall (the seed treatment model) were found for number of workers and number of filled storage cells (pollen and nectar/honey cells). Significant effects were also found in the additive seed treatment model (treatment*country) for number of workers, eggs and filled storage cells. All three of these additive models were found to be a better fit to the data than the overall seed treatment model, suggesting that while there was an effect of seed treatment, this differed by country.
No overall effect of seed treatment was found on the overwintering performance (in Germany & Hungary, UK overwintering mortality was too high to allow statistical analysis). The additive model showed significantly lower overwintering worker numbers in Hungary vs Germany.
Bumblebees and solitary bees
The same modelling approach (seed treatment vs null, additive (treatment*country) vs null, seed treatment vs additive) was carried out for the wild bees. The models included country, the percentage cover of OSR around each site, percentage cover of arable land around each site, the maximum neonic residue concentration (natural logs) in both colony stores and in the crop, and the median neonic residue concentration (natural log) in colony stores as covariates, as these were found to explain some of the variation between sites.
No overall (seed treatment only) effect was found: neither Buff-tail bumbles (colony weight gain, queen/male/worker production) nor Red Mason bees (number of reproductive cells produced) were significantly affected merely by living in areas treated with pesticides.
When examined at a country level (the additive model), in the UK the production of Buff-tail males was significantly lower in treated fields than in controls, but higher in treated fields than in controls in Germany.
Overall across all three countries combined, production of new reproductive individuals (Buff-tail bumblebee queen production, Red Mason bees reproductive cell production) significantly fell and correlated with both the peak and median levels of neonic residues in the nests (clothianidin, thiamethoxam & imidocloprid combined & corrected for relative toxicity to Honey bees), strongly suggesting a causal link. When the contaminant imidocloprid residues were removed from the analysis, there remained a significant correlation between Buff-tail bumblebee queen production and combined clothianidin and thiamethoxam residues, but the correlation for Red Mason bees became insignificant.
The CEH paper is thorough and well-carried-out – particularly when taken as a study of real-world usage of neonic seed treatments rather than of the neonicotinoid chemicals alone. For honeybees, the results are ambiguous: of the 30 tests of hive health during the OSR field season, only five were significant, and of these two had a positive effect on hive health (thiamethoxam on UK storage cells and clothianidin on German egg cells) and three were negative (clothianidin on UK worker numbers, UK storage cells, and Hungarian egg cells). Of the 16 tests of overwintering hive health, only one was significant: the negative effect of clothianidin on Hungarian worker numbers.
The likely reason for this is that there are several other factors in play (as the text points out: ‘the effects of neonicotinoids are a product of interacting factors’). While neonics (each of which are different in their effects on bees) do have detrimental sub-lethal effects on Honey bees, these are largely at the individual level. Prolonged, constant chronic exposure is required for effects to become visible at the hive level (Tsvetkov et al. 2017), and effects are likely to be moderated by overall hive stress levels. These are in turn affected by foraging breadth, pest/pathogen load, weather effects, etc, all of which were different between countries here. Previous work has also found a correlation between the strength of a colony at initiation and the development rate (Schmuck and Lewis 2016). The paper concludes ‘our results suggest that exposure to low levels of neonicotinoids may cause reductions in hive fitness that are influenced by a number of interacting environmental factors’ and overall it seems likely that under field conditions, effects of clothianidin and thiamethoxam on colony health are difficult to discern from the statistical background noise.
For wild bees, the results are more damning. Although treatment type had no effect (considered in isolation), or mixed effects (considered per country: increased male production in Germany, but decreased male production in the UK for Buff-tail bumblebees), this is unsurprising as the crops were generally not significantly different in terms of neonic production. However, when the production of reproductive individuals (queens Buff-tail bumblebees, reproductive cells of Red Mason bees) are modelled against their direct exposure to neonics (the residue found inside the nest (Buff-tail bumblebees) or brood cells (Red Mason bees)), significant correlation is found.
This backs up and extends the results of the previous large-scale field study on wild bees, which found significant effects on reproduction in Buff-tail bumblebees and Red Mason bees (Rundlöf et al. 2015), as well as lab studies which found sublethal effects. This study provides strong evidence for chronic negative effects of exposure to neonics on the reproductive capacity of wild bees.