The environmental risks of neonicotinoid pesticides: a review of the evidence post 2013

Since 2014, a number of studies have been published which report neonicotinoid concentrations in the pollen and nectar of neonicotinoid-treated flowering crops. These results have been approximately in line with the concentrations reported by EFSA and Godfray et al. In oilseed rape treated with thiamethoxam, Botías et al. (2015) found average concentrations of 3.26 ng/g of thiamethoxam, 2.27 ng/g of clothianidin and 1.68 ng/g of thiacloprid in the pollen. Oilseed rape nectar contained similar average concentrations of 3.20 ng/g of thiamethoxam, 2.18 ng/g of clothianidin and 0.26 ng/g of thiacloprid. Xu et al. (2016) found average levels of clothianidin in oilseed rape of 0.6 ng/g. No pollen samples were taken. In maize pollen, Stewart et al. (2014) found average thiamethoxam and clothianidin levels between the limit of detection (LOD) of 1 to 5.9 ng/g across a range of seed treatments. Xu et al. (2016) found average clothianidin concentration of 1.8 ng/g in maize pollen. Additionally, Stewart et al. (2014) found no neonicotinoid residues in soybean flowers or cotton nectar.

This level of variation of up to one order of magnitude in neonicotinoid concentrations found in bee-collected pollen and nectar in different studies is substantial. The detected levels in pollen and nectar presumably depend significantly on the dose and mode of treatment, the studied crop, the season, the location, the soil type, the weather, time of day samples are collected, and so on. Even different crop varieties can result in significant variation in the residue content of pollen and nectar (Bonmatin et al. 2015). Because pollen samples taken from a series of bees will be from a mixture of different plants, most of which will not be crop plants, the neonicotinoid residues in crop pollen will be diluted by untreated, non-crop pollen. However, for the reported studies, the higher neonicotinoid concentrations are within an order of magnitude of the 6.1 ng/g in pollen and 1.9 ng/mL in nectar values calculated by Godfray et al. (2014). Additionally, in all cases, the concentrations of neonicotinoids in pollen and nectar were higher at sites adjacent to neonicotinoid-treated flowering crops than at sites adjacent to untreated crops. The available evidence shows that proximity to treated flowering crops increases the exposure of bees to neonicotinoid pesticides. The recent evidence for concentrations found in flowering crops is approximately in line with the levels reported by EFSA (2013a, b, c).

The EFSA studies state that some of the crops on which clothianidin is authorised as a seed-dressing do not flower, are harvested before flowering or do not produce nectar or pollen, and therefore, these crops will not pose any risk to bees via this route of exposure. Whilst non-flowering crops are clearly not a source of exposure through produced pollen and nectar, they do represent a source of neonicotinoids that can dissipate into the wider environment (discussed in the “Risk of exposure for non-target organisms from neonicotinoids persisting in the wider environment” section). Additionally, treated crops of any type represent additional pathways of neonicotinoid exposure to other organisms.

Depending on crop species and consequent seed size, neonicotinoid-treated seeds contain between 0.2 and 1.25 mg of active ingredient per seed (Goulson 2013). For a granivorous grey partridge weighing 390 g, based on typical treatment rates, Goulson calculated that it would need to consume around five maize seeds, six sugar beet seeds or 32 oilseed rape seeds to receive a nominal LD₅₀. Based on US Environmental Protection Agency (EPA) estimates that around 1% of sown seed is accessible to foraging vertebrates at recommended sowing densities, Goulson calculated that sufficient accessible treated seed would be present to deliver a LD₅₀ to ∼100 partridges per hectare sown with maize or oilseed rape. Given that grey partridges typically consume around 25 g of seed a day, there is the clear potential for ingestion of neonicotinoids by granivorous animals, specifically birds and mammals. However, whilst some experimental studies have been conducted to investigate mortality and sublethal effects of treated seeds on birds (see the “Sensitivity of birds and bats to neonicotinoids” section), no studies are available that demonstrate consumption of treated seed by farmland birds under field conditions or quantify relative consumption of treated versus untreated seed to better understand total exposure via this route.

In addition to insect herbivores, developing seedlings treated with neonicotinoids are predated by molluscan herbivores which can be serious pests of arable crops. Because neonicotinoids have relatively low efficacy against molluscs, Douglas et al. (2015) investigated neonicotinoid residues in the slug Deroceras reticulatum, a major agricultural pest, using neonicotinoid seed-treated soybean in both laboratory and field studies. Total neonicotinoid concentrations from samples of field collected slugs feeding on treated soybean were as high as 500 ng/g with average levels over 100 ng/g after 12 days of feeding. No neonicotinoids were detected in slugs feeding on untreated control plants. After 169 days, no neonicotinoids were detected in either control or treated slugs. In the laboratory, slugs consuming soybean seedlings incurred low mortality of between 6 and 15% depending on the strength of the seed treatment. In laboratory experiments, slugs were exposed to the ground beetle Chlaenius tricolour after feeding on soybean. C. tricolour is a typical predatory beetle found in agro-ecosystems and is known to be an important predator of slugs. For beetles that consumed slugs, 61.5% (n = 16/26) of those from the neonicotinoid treatment subsequently showed signs of impairment compared to none of those in the control treatment (n = 0/28). Of the 16 that showed impairment, seven subsequently died. A similar result was found by Szczepaniec et al. (2011) who found that the application of imidacloprid to elm trees caused an outbreak of spider mites Tetranychus schoenei. This increase was as a result of a reduction in the density of their predators which incurred increased mortality after ingesting imidacloprid-containing prey items. Many beneficial predatory invertebrates feed on pests of crops known to be treated with neonicotinoids, but to date no other studies have assessed whether neonicotinoids are transmitted to these predators through direct consumption of crop pests in agro-ecosystems.

Additionally, flowering crops in a non-flowering stage can also pose a potential threat to natural enemy populations. The soybean aphid parasitoid wasp Aphelinus certus is an important parasite of the soybean aphid Aphis glycines. Frewin et al. (2014) gave A. certus access to laboratory populations of aphids feeding on control and neonicotinoid-treated soybean plants. A. certus parasitised a significantly smaller proportion of aphids on treated plants than on untreated plants. Frewin et al. hypothesise two potential reasons for this effect—firstly that exposure to neonicotinoid residues within aphid hosts may have increased mortality of the immature parasitoid or the parasitism combined with residues may have increased aphid mortality. Secondly, A. certus may avoid parasitising pesticide-poisoned aphids. Aphelinus species are known to use internal cues to determine host suitability, and it is possible that they may use stress- or immune-related aphid hormones to judge host suitability. Given that a key part of biological control of insect pests using parasitic wasps is to increase the parasitoid abundance early in the season, the reduction in the parasitism rate caused by neonicotinoid seed treatment could potentially impair the ability of A. certus to control soybean aphid. It is not known if A. certus emerging from contaminated hosts will incur lethal or sublethal effects which may further impair this ability.

Non-flowering neonicotinoid crops present possible exposure routes through direct consumption of treated seed or consumption of seedling plants that may result in the transmission of neonicotinoids to higher trophic levels, including beneficial insects that offer a level of pest control through predatory behaviour. As the EFSA reports did not consider the impact of neonicotinoids on non-bees, no comparison can be made here.

Numerous studies (12 listed by Godfray et al. 2014) prior to 2013 identified that neonicotinoids present in seed dressings can be mechanically abraded during the drilling process and can subsequently be emitted as dust. This dust can contain very high levels of neonicotinoids, up to 240,000 ng/g under certain conditions (see the review by Nuyttens et al. 2013). Acute contact with this dust can in certain cases result in the mass poisoning of honeybees (e.g. Pistorius et al. 2009; Bortolotti et al. 2009). Concentrations of neonicotinoids in dust created during sowing and the total volume released into the air depend on application rate, seed type, seed treatment quality (including additions of seed lubricants such as talcum powder), seed drilling technology and environmental conditions. Girolami et al. (2013) demonstrated that the dust cloud created by seed drills is an ellipsoidal shape approximately 20 m in diameter. Using cage experiments, a single pass of a drilling machine was sufficient to kill all honeybees present. The use of tubes designed to direct exhaust air towards the ground did not substantially increase bee survival rate. Neonicotinoid concentrations of up to 4000 ng/g were detected in honeybees with an average concentration of 300 ng/g. Similar concentrations were detected in bees exposed to both unmodified and modified drills.

On the basis of the available evidence, the EFSA reports (2013a, b, c) concluded that maize produces the highest dust drift deposition, whilst for sugar beet, oilseed rape and barley seeds the dust drift deposition was very limited. No information was available for other crops, and given that seed type is an important factor determining neonicotinoid release, extrapolation to other crops is highly uncertain. A high acute risk was not excluded for bees foraging or flying in adjacent crops during the sowing of maize, oilseed rape and cereals. In practice, this assessment indicates that forager honeybees or other pollinators flying adjacent to the crop are at high risk (e.g. via direct contact to dust) and may be able to carry considerable residues back to the hive (for social bees). Bees present further away or foraging upwind during the sowing will be considerably less exposed. The reports conclude that the aforementioned assessments do not assess potential risk to honeybees from sublethal effects of dust exposure. No information on neonicotinoid residues in nectar in the adjacent vegetation following dust drift was available.

In recent years, various types of improved seed drills have been adopted that direct air from the drills towards the soil, reducing the dust drift effect by up to 95% (see Manzone et al. 2015). Air deflectors have become mandatory for certain products in the Netherlands, France, Belgium and Germany (Godfray et al. 2014). Bonmatin et al. (2015) and Long and Krupke (2015) reviewed existing literature on the exposure of pollinators and other non-target organisms to contaminated dust from seed drilling machines. The authors conclude that despite attention by regulators they consider dust drift to be a likely cause of environmental neonicotinoid contamination, in particular when best practice is not followed.

Recent studies continue to detect neonicotinoids in the tissues of wildflowers surrounding agricultural fields immediately after planting. Stewart et al. (2014) detected average neonicotinoid concentrations of 9.6 ng/g in whole wildflowers collected from field margins adjacent to fields planted with maize (n = 18), cotton (n = 18) and soybean (n = 13). The samples were collected a few days after sowing (typically within 3 days), with the highest concentration of 257 ng/g collected adjacent to a maize field sown the previous day with thiamethoxam-treated seed. Detailed data on concentrations adjacent to each crop type are not available. No samples were taken from vegetation adjacent to crops sown without a neonicotinoid seed dressing. Rundlöf et al. (2015) collected flowers and leaves from wild plants growing adjacent to treated and untreated oilseed rape fields 2 days after sowing. Adjacent to the treated fields, neonicotinoid concentrations were lower than in the previous study at 1.2 ng/g, but this was higher than the control fields where no neonicotinoids were detected.

Some plants secrete small volumes of liquid (xylem sap) at the tips of leaves or other marginal areas, often referred to as guttation droplets. Six published studies and an EFSA review found extremely high neonicotinoid concentrations in guttation droplets of up to four to five orders of magnitude greater than those found in nectar, particularly when plants are young (see Godfray et al. 2014). Using a clothianidin concentration of 717,000 ng/g and an acute oral toxicity of 3.8 ng/bee for clothianidin (see the “Direct lethality of neonicotinoids to adult wild bees” section), EFSA (2013a) calculated that a honeybee would only need to consume 0.005 μl to receive an LD₅₀. Given that honeybee workers can carry between 1.4 and 2.7 mL of water a day, there is the clear potential for lethal exposure via this route. The risk assessments for thiamethoxam and imidacloprid were similar (EFSA 2013b, c). However, on the basis of experimental trials, the EFSA reports conclude that whilst guttation droplets were frequently produced, honeybees were rarely seen collecting water from them and therefore the risk should be considered low.

Few studies have looked at neonicotinoid exposure via guttation droplets since 2013. In the one available study, Reetz et al. (2015) assessed thiamethoxam concentrations in oilseed rape guttation droplets and measured residues in individual honeybee honey sacs. The authors note that targeted observations of water-foraging honeybees in the field are nearly impossible, and so returning honeybees from apiaries placed out adjacent to treated oilseed rape crops were instead collected in the autumns of 2010 and 2011 when seedling oilseed rape crops were producing guttation droplets. Oilseed rape produced guttation droplets containing between 70 and 130 ng/mL clothianidin at the cotyledon stage. Out of 436 honey sacs, neonicotinoids were only detected in 62 samples at concentrations between 0.1 and 0.95 ng/mL. However, because there was no behavioural observation, it is not possible to state the origin of this contamination with certainty; neonicotinoids are also present in waterbodies and the nectar of wild flowers (see the “Risk of exposure for non-target organisms from neonicotinoids persisting in the wider environment” section). As such, there is still little evidence documenting the extent to which honeybees or other insects collect or are otherwise exposed to neonicotinoids through contact with guttation droplets.

Although neonicotinoids applied through a seed dressing are designed to be taken up into the target crop plant, only 1.6–20% of the active ingredient is absorbed, with the majority remaining in the soil (Sur and Stork 2003; Goulson 2014; Bonmatin et al. 2015). A small proportion is dispersed through dust created whilst drilling (see the “Risk from non-flowering crops and cropping stages prior to flowering” section). Neonicotinoids can bind to soil with the strength of the binding dependent on various factors. Neonicotinoids are water soluble (see the “Persistence of neonicotinoids in water and transport mechanisms for contamination of aquatic systems” section) and may leach from soils if water is present. Leaching is lower and sorption is higher in soils with a high content of organic material (Selim et al. 2010). In a recent comparison of soil types, Mörtl et al. (2016), Fig. 3) found that clothianidin and thiamethoxam leached readily from sandy soils. Clay soils showed higher retention of neonicotinoids, but the greatest retention was seen for loam soils. Correspondingly, the highest residual neonicotinoid concentrations were found in loam soils.

Open image in new window

Whilst several studies have assessed dissipation half-life times (DT₅₀) of neonicotinoids in soil, much of this work was conducted before the recent interest in the potentially deleterious effect of neonicotinoids on wider biodiversity. A review of available DT₅₀ times from field and laboratory studies conducted between 1999 and 2013 was reviewed by Goulson (2013). Reported DT₅₀s are highly variable and typically range from 200 to in excess of 1000 days for imidacloprid, 7–353 days for thiamethoxam and 148–6931 days for clothianidin. DT₅₀s appear to be shorter for the nitro-substituted neonicotinoids, at 3–74 days for thiacloprid and 31–450 days for acetamiprid. DT₅₀ values of over 1 year would suggest the likelihood of neonicotinoid accumulation in the soil, assuming continuous input. However, these reported values are highly variable. At the time the EFSA reports were written, only one field study was available that assessed neonicotinoid accumulation in the soil over multiple years with continued neonicotinoid input. Bonmatin et al. 2005 screened 74 samples of farmland soil from France for imidacloprid. Imidacloprid concentrations were higher in soils which had been treated in two consecutive years than those soils which had only received one treatment, suggesting the possibility of imidacloprid accumulation in the soil. However, as the study only looked at soils treated for a maximum of 2 years, it is not clear whether residues would continue to increase. Two studies had been completed by 2013 but were not widely disseminated. These studies were carried out by Bayer and assessed levels of imidacloprid in soil over 6 years for seed-treated barley in the UK (Placke 1998a) and spray application to orchard soils in Germany (Placke 1998b). Goulson (2013) reviewed this data and argued that the studies show accumulation of neonicotinoids in soils over time, with some indication that concentrations may begin to plateau after about 5 years. However, since the trials were terminated after 6 years, it is not clear whether levels would have continued to increase.

Botías et al. (2015) analysed soil samples from seven winter-sown oilseed rape and five winter-sown wheat fields collected in summer 2013, 10 months after the crops were sown. Samples were collected from field centres (oilseed rape only) and field margins (oilseed rape and winter wheat). Imidacloprid (range ≤0.07–7.90 ng/g), clothianidin (range 0.41–28.6 ng/g), thiamethoxam (range ≤0.04–9.75 ng/g) and thiacloprid (range ≤0.01–0.22 ng/g) were detected. Residues from the centre of the oilseed rape fields were higher than for the edge of the oilseed rape fields (average imidacloprid 3.03 against 1.92 ng/g, average clothianidin 13.28 against 6.57 ng/g, average thiamethoxam 3.46 against 0.72 ng/g and average thiacloprid 0.04 against ≤0.01 ng/g). Whilst these values are higher than those measured by Jones et al. (2014) and Limay-Rios et al. (2015), they are within an order of magnitude at their greatest difference.

Hilton et al. (2015) presented previously private data from 18 industry trials conducted between 1995 and 1998 for thiamethoxam applied to bare soils, grass and a range of crops (potatoes, peas, spring barley, winter barley, soybean, winter wheat and maize). Thiamethoxam DT₅₀s ranged between 7.1 and 92.3 days, with a geometric mean of 31.2 days (arithmetic mean 37.2 days). Across different application methods and environmental conditions, thiamethoxam declined to <10% of its initial concentration within 1 year. de Perre et al. (2015) measured soil clothianidin concentrations over 2011 to 2013, with clothianidin-treated maize sown in the springs of 2011 and 2013. Maize seeds were sown with seed dressings of 0.25 and 0.50 mg/seed (Fig. 4). At the lower-concentration seed dressing, clothianidin residues in the soil ranged from approximately 2 ng/g before planting to 6 ng/g shortly after planting. At the higher seed dressing, clothianidin average residues ranged from 2 ng/g before planting to 11.2 ng/g shortly after planting. For the seed treatment of 0.5 mg/seed, de Perre et al. (2015) calculated a DT₅₀ for clothianidin of 164 days. For the lower treatment of 0.25 mg/seed, a DT₅₀ of 955 days was calculated, though this model explained a much lower proportion of the data than the model for the 0.5 mg/seed data.

Open image in new window

Schaafsma et al. (2016) calculated clothianidin DT₅₀s in maize fields in Ontario, Canada, in 2013 and 2014, including data published in Schaafsma et al. (2015). Soil samples were collected from 18 fields in the spring before crop planting. Average neonicotinoid concentrations (clothianidin and thiamethoxam aggregated) were 4.0 ng/g in 2013 and 5.6 ng/g in 2014. Using the observed residues and the recharge rate applied at planting via treated maize seeds, fields studied in 2013 had an estimated DT₅₀ of 0.64 years (234 days) and fields studied in 2014 had an estimated DT₅₀ of 0.57 years (208 days). For fields studied in both years, DT₅₀ was calculated at 0.41 years (150 days). Schaafsma et al. conclude that, at current rates of neonicotinoid application in Canadian maize cultivation, soil residues of neonicotinoids will plateau at under 6 ng/g.

Xu et al. (2016) analysed soil samples from 50 maize-producing sites in the Midwestern USA across 2012 and 2013 and soil samples from 27 oilseed rape-producing sites in western Canada across 2012, 2013 and 2014. Samples were collected after planting, but it is not clear exactly how long after. Average clothianidin soil concentration at Midwestern maize-producing sites with a range of 2–11 years of planting clothianidin-treated seeds was 7.0 ng/g with a 90th percentile concentration of 13.5 ng/g. Xu et al. argue that this average is similar to the theoretical soil concentrations (6.3 ng/g) expected from a single application of 0.25 mg clothianidin-treated maize seed. Clothianidin levels in soil appear to plateau after 4 years, but the sample size for sites with a history of more than 4 years is much smaller than the number of sites with a history of under 4 years of use. At the oilseed rape-producing sites, average clothianidin concentrations were 5.7 ng/g with the 90th percentile concentration of 10.2 ng/g. This is also similar to the theoretical soil concentration (6.7 ng/g) from a single application of oilseed rape seed treated at 4 g clothianidin per kilogram of seed. The oilseed rape sites do not have the same history of clothianidin use, but levels appear to be fairly stable over the 4 years of applications. For reference, 10 g clothianidin per kilogram of oilseed rape seed is the most common dosage rate in recent field trials (the Elado seed dressing, the “Impact on colony growth and reproductive success” section).

The current body of evidence shows that detectable levels of neonicotinoids are found in agricultural soils over a year after treated seeds were planted, clearly demonstrating a level of neonicotinoid persistence greater than the annual agricultural cycle. Moreover, neonicotinoids known not to have been recently used can still be present in soils several years after the last application date. The available data suggest that, whilst a proportion of the total neonicotinoids applied can and do persist in the soil from year to year, there appears to be sufficient degradation that means they do not continue to accumulate indefinitely but instead plateau after 2–6 years of repeated application. However, these studies also show that overall, the annual sowing of neonicotinoid-treated seed results in chronic levels of neonicotinoid soil contamination in the range of 3.5–13.3 ng/g for clothianidin and 0.4–4.0 ng/g for thiamethoxam which will act as a constant source of exposure for soil-dwelling organisms, and for neonicotinoid transport into the wider environment.

Neonicotinoids are soluble in water, a property that is necessary for them to function effectively as systemic pesticides which can be taken up by crops. The solubility of neonicotinoids depends on local conditions such as ambient temperature, water pH and the form that the neonicotinoids are applied in, such as granules, as a seed dressing or as dust drift from seed drilling (Bonmatin et al. 2015). Under standard conditions (20 °C, pH 7), neonicotinoid solubility varies between 184 mg/L (moderate) to 590,000 mg/L (high) for thiacloprid and nitenpyram respectively (PPDB 2012). The values for clothianidin, imidacloprid and thiamethoxam are 340 mg/L (moderate), 610 mg/L (high) and 4100 mg/L (high) respectively. In contrast, Fipronil has a solubility two to three orders of magnitude lower at 3.78 mg/L under the same conditions.

Because of the high solubility of neonicotinoids in water, concerns were raised that neonicotinoids might be passing into waterbodies in the wider environment and that this may pose a risk for aquatic organisms. Available evidence to 2015 was reviewed by Bonmatin et al. 2015 and Morrissey et al. 2015. In general, under simulated environmental conditions, neonicotinoids readily leach into water (Gupta et al. 2008; Tišler et al. 2009). Neonicotinoids have been identified passing into waterways through several different routes. These include direct leaching into groundwater and subsequent discharge into surface water, decay of treated plant material in waterways and direct contact from dust from the drilling of treated seed, treated seeds or spray drift into waterbodies (Krupke et al. 2012; Nuyttens et al. 2013). The majority of this contamination is thought to occur from run-off after acute rainfall (Hladik et al. 2014; Sánchez-Bayo and Hyne 2014; Main et al. 2016). Run-off will be particularly severe where soil organic content is low and on steep slopes (Goulson 2013).

Whilst rainfall during or shortly after the planting season appears to be the main mechanism for neonicotinoid transport into waterbodies, detectable levels of neonicotinoids can be found in prairie wetlands in Canada during early spring before the planting season (Main et al. 2014). Main et al. (2016) analysed snow, spring meltwater, particulate matter and wetland water from 16 wetland sites adjacent to agricultural fields that had been used to grow either oilseed rape (canola, treated with neonicotinoids) or oats (not treated). They found that all meltwater samples were contaminated with clothianidin and thiamethoxam in the range of 0.014–0.633 μg/L (1 μg/l = 1 ppb). Levels of contamination in meltwater were higher adjacent to fields planted with neonicotinoid-treated oilseed rape in the previous year (mean 0.267 μg/l). However, fields planted with non-neonicotinoid-treated oats in the previous year still showed similar levels of contamination (mean 0.181 μg/l). Treated oilseed rape and untreated oats are frequently rotated from year to year (Main et al. 2014), and the small difference in neonicotinoid concentration in meltwater from fields previously planted with treated and untreated crops suggests the persistence of neonicotinoids in the soil over multiple years (see the “Persistence of neonicotinoids in water and transport mechanisms for contamination of aquatic systems” section). The findings of this study suggest that neonicotinoid active ingredients previously bound to soil particles are eroded during spring freeze-thaw cycles. The demonstration of this route of transport in addition to general rainfall suggests a more chronic transport of neonicotinoids into waterbodies outside the main period of crop planting.

In addition to these peer-reviewed studies, Lu et al. drew comparison with European Commission regulatory studies on neonicotinoid compounds (European Commission (EC) 2004a, b, 2005, 2006). The European Commission studies found half-lives in water of 3.3 h for clothianidin, 2.3–3.1 days for thiamethoxam, 34 days for acetamiprid and 80 days for thiacloprid. The exact methodology used in these studies is unclear and inconsistent (see Lu et al. 2015 discussion). Nevertheless, the overall trend is consistent with the cyano-substituted neonicotinoids (acetamiprid and thiacloprid) taking one to two orders of magnitude longer to degrade than the nitro-substituted neonicotinoids (thiamethoxam, clothianidin and imidacloprid). The short half-lives of these three, most widely used neonicotinoids suggests that, under field conditions, free neonicotinoids in surface waters should be broken down by natural light in a matter of hours or days. However, local environmental conditions can affect this, with increasing turbidity increasing neonicotinoid persistence. Moreover, in mesocosm experiments, photolysis of thiamethoxam was found to be negligible at depths of greater than 8 cm (Lu et al. 2015). This significant light attenuation through the water column suggests that neonicotinoids may be shielded from photolysis even in shallow waterbodies. In waterbodies such as groundwater that are not exposed to light, there will be no photolysis. In these circumstances, clothianidin is persistent and has the potential to accumulate over time (Anderson et al. 2015), though empirical data demonstrating this is lacking.

The most comprehensive review of levels of neonicotinoid contamination in global surface waters was conducted by Morrissey et al. (2015), though see also Anderson et al. (2015). Morrissey reviewed reported average and peak levels of neonicotinoid contamination from 29 studies from nine countries between 1998 and 2013. The waterbodies studied included streams, rivers, drainage, ditches, groundwater, wetlands, ponds, lakes, puddled surface waters and run-off waters. Study systems were adjacent to or receiving run-off water from agricultural land. From this dataset (Fig. 5), the geometric mean for average surface water neonicotinoid concentration was 0.13 μg/l (=0.13 ppb, n = 19 studies) and the geometric mean for peak surface water concentration was 0.63 μg/l (=0.63 ppb, n = 27 studies). Because most monitoring schemes use spot sampling, they are likely to underreport the true maximum concentrations that occur immediately after maximum periods of neonicotinoid influx (Xing et al. 2013). As peak concentrations are often found after acute events such as heavy rainfall, this limits our understanding of the true average and maximum concentrations that are found in waterbodies.

Open image in new window

Since Morrissey et al. (2015) was published, a number of studies have become available documenting broadly similar neonicotinoid contamination levels in a wide range of aquatic environments. At a small scale in agricultural regions, Schaafsma et al. (2015) measured concentrations in surface water (puddles and ditches) in and around 18 maize fields in Ontario, Canada. They found arithmetic mean residues of 0.002 μg/L of clothianidin (maximum = 0.043 μg/L) and 0.001 μg/L of thiamethoxam (maximum = 0.017 μg/L). In Iowa, USA, Smalling et al. (2015) assessed six wetlands surrounded by agricultural land and found arithmetic mean neonicotinoid concentrations of 0.007 μg/L (maximum 0.070 μg/L). Away from agricultural land, Benton et al. (2016) measured concentrations in mountain streams in the southern Appalachians, USA, where eastern hemlock forests are treated with imidacloprid to control pests. Average concentrations of 0.067 μg/L of imidacloprid (maximum = 0.379 μg/L) were found in seven of the 10 streams investigated. de Perre et al. (2015) measured concentrations of clothianidin in groundwater below fields of treated maize. Data on average concentrations are not available, but concentrations peaked at 0.060 μg/L shortly after crop planting.

At a wider scale, Qi et al. (2015) and Sadaria et al. (2016) measured concentrations in wastewater treatment plants. Qi et al. (2015) recorded imidacloprid at concentrations between 0.045 and 0.100 μg/L in influent and 0.045 and 0.106 μg/L in effluent at five wastewater treatment plants in Beijing, China, with no data available on arithmetic mean concentrations. Sadaria et al. (2016) assessed influent and effluent wastewater at 13 conventional wastewater treatment plants around the USA. For influent, imidacloprid was found at arithmetic mean concentrations of 0.061 μg/L, acetamiprid at 0.003 μg/L and clothianidin at 0.149 μg/L. For effluent, imidacloprid was found at concentrations of 0.059 μg/L, acetamiprid at 0.002 μg/L and clothianidin at 0.070 μg/L.

Two nationwide surveys for neonicotinoids were also published. Hladik and Kolpin (2016) measured neonicotinoid concentrations in 38 streams from 24 US states plus Puerto Rico. Five neonicotinoids (acetamiprid, clothianidin, dinotefuran, imidacloprid, thiamethoxam) were recorded with at least one compound found in 53% of sampled streams, with an arithmetic mean contamination of 0.030 μg/L and median contamination of 0.031 μg/L. Thiacloprid was not recorded. Székács et al. (2015) conducted a nationwide survey of Hungarian watercourses, finding clothianidin at concentrations of 0.017–0.040 μg/L and thiamethoxam at concentrations of 0.004–0.030 μg/L.

Across all studies, the highest levels of neonicotinoid contamination were found in agricultural areas. In the most comprehensive nationwide survey of streams across the USA conducted between 2012 and 2014, levels of clothianidin and thiamethoxam contamination (the now dominant agricultural neonicotinoids) were significantly positively correlated with the proportion of the surrounding landscape used for crop cultivation (Hladik and Kolpin 2016). The most acute levels of neonicotinoid contamination in agricultural areas are reported from surface water in the immediate vicinity of cultivated crops. Puddles adjacent to fields planted with neonicotinoid-treated maize seeds were found to contain maximum concentrations of 55.7 μg/L clothianidin and 63.4 μg/L thiamethoxam in Quebec, Canada (Samson-Robert et al. 2014). Surface water in the Netherlands had imidacloprid concentrations up to 320 μg/L (van Dijk et al. 2013), and transient wetlands found in intensively farmed areas of Texas had thiamethoxam and acetamiprid concentrations of up to 225 μg/L (Anderson et al. 2013). In Hungary, the highest neonicotinoid concentrations of 10–41 μg/L were found in temporary shallow waterbodies after rain events in early summer (Székács et al. 2015). More generally, watercourses draining agricultural fields had high levels of neonicotinoids after rainfall in Canada, the USA and Australia (Hladik et al. 2014, Sánchez-Bayo and Hyne 2014). Where repeated sampling of the same site has been carried out, the highest neonicotinoid concentrations have been found in early summer and are associated with rainfall during the planting season (Main et al. 2014; Hladik et al. 2014). Hladik and Kolpin (2016) measured neonicotinoid concentrations in three agriculturally affected streams in Maryland and Pennsylvania and found peak levels after rain events during the crop planting season in May, though this could not be formally statistically analysed due to low sample size (Fig. 6).

Open image in new window

In addition to agricultural run-off, urban areas also contribute towards neonicotinoid contamination of waterbodies. Whilst the use of imidacloprid as an agricultural pesticide has declined, it is still found in a wide range of domestic products and veterinary treatments for pets (Goulson 2013). Hladik and Kolpin (2016) continuously monitored neonicotinoid levels in Slope Creek, a stream surrounded by a largely urban catchment (39% urban), and the Chattahoochee River which includes the drainage of Slope Creek and overall has a lower proportion of urbanisation (9%). Imidacloprid was the dominant neonicotinoid found, present in 87% of the 67 collected samples (Fig. 7). Dinotefuran and acetamiprid were less frequently encountered. Unlike in the studied watercourses draining agricultural land, no significant relationship was seen with stream flow in either Slope Creek or the Chattahoochee River. Hladik and Kolpin suggest that this may be because, unlike for the planting period of arable crops, there is no distinct period of use for domestic imidacloprid in an urbanised catchment. No clothianidin or thiamethoxam was detected, probably because neither catchment contained cultivated crops.

Open image in new window

Botías et al. (2015) collected pollen and nectar from wildflowers growing in field margins adjacent to agricultural fields planted with neonicotinoid-treated oilseed rape and wheat. Pollen samples from 54 wild flower species were collected. Thiamethoxam, imidacloprid and thiacloprid were all detected. Thiamethoxam was the most frequently encountered neonicotinoid, and levels were highly variable with the highest concentrations found in Heracleum sphondylium at 86 ng/g and Papaver rhoeas at 64 ng/g. There was substantial variation in the levels of contamination in the same wildflower species found in different field margins. Average levels of total neonicotinoid contamination in wildflower pollen were significantly higher in margins adjacent to treated oilseed rape (c. 15 ng/g) than for margins adjacent to treated wheat (c. 0.3 ng/g). Levels of neonicotinoids were much lower in wild plant nectar. Only thiamethoxam was detected at average levels of 0.1 ng/g in wild flowers adjacent to oilseed rape fields and <0.1 ng/g adjacent to wheat fields.

Botías et al. (2015) is the only available study which has specifically measured neonicotinoid concentrations in pollen and nectar directly taken from wild plants growing in close proximity to neonicotinoid-treated crops. Mogren and Lundgren (2016) assessed neonicotinoid concentrations in the nectar of five wild flower species sown as part of pollinator conservation measures which were located adjacent to neonicotinoid-treated maize. This was achieved by collecting honeybees seen to visit these flowers for nectar and extracting the contents of their crop for neonicotinoid residue analysis. Honeybees generally have a very high fidelity to visiting the same flower species on a single forage flight so the authors assumed that the nectar was representative of that particular species. Average clothianidin concentrations found in this nectar ranged between 0.2 and 1.5 ng/g, with significant differences found between wild plant species. Mogren and Lundgren (2016) also tested the foliage of seven wildflower species for neonicotinoid residues directly. There was high variability in clothianidin uptake between and within plant species. Sunflowers Helianthus annuus accumulated the highest levels with concentrations of 0–81 ng/g, with buckwheat Fagopyrum esculentum and phacelia Phacelia tanacetifolia accumulating lower levels at 0–52 and 0–33 ng/g respectively. Similarly, high levels of variation were found by Botías et al. (2016) who sampled the foliage of 45 species of wild plant in field margins adjacent to treated oilseed rape crops. Average total neonicotinoid contamination was 10 ng/g, with the highest levels seen in creeping thistle Cirsium arvense of 106 ng/g of thiamethoxam. Pecenka and Lundgren (2015) looked specifically at clothianidin concentrations in milkweed Asclepias syriaca in field margins adjacent to clothianidin-treated maize. Levels were lower than the previous two studies, with mean levels of 0.58 ng/g with a maximum concentration of 4.02 ng/g.

Whilst not looking at specific concentrations in pollen, nectar or foliage, Stewart et al. (2014) and Rundlöf et al. (2015) found total mean neonicotinoid concentrations of 10 and 1 ng/g respectively in whole wild flower samples collected around neonicotinoid-treated fields. As discussed in the “Risk of exposure from the drilling of treated seed and subsequent dust drift” section, these levels may have been a direct result of neonicotinoid-contaminated dust drift onto surrounding vegetation and do not in and of themselves demonstrate uptake of neonicotinoids from contaminated soil and/or water.

Across all studies published since 2013, average levels of neonicotinoids in wild plants range from 1.0 to 7.2 ng/g in whole flower samples, 0.4 to 13.5 ng/g in foliage samples, <0.1 to 1.5 ng/g in nectar samples and <0.04 to 14.8 ng/g in pollen samples. Due to the limited number of studies available, it is difficult to make a comparison with levels in directly treated crop plants. However, they are broadly comparable to the levels found in the treated crop itself (see the “Risk of exposure from pollen and nectar of treated flowering crops” section).

In 2013, it was known that honeybees collected neonicotinoid contaminated pollen from crop plants, but the extent to which this was diluted by uncontaminated pollen from wild plants was unknown. Krupke et al. (2012) found levels of clothianidin and thiamethoxam in honeybee-collected pollen that ranged between 0 and 88 ng/g, with the proportion of pollen collected from maize (the main treated crop in their study area) also varying substantially between 2.6 and 82.7%. There was no correlation between the proportion of maize pollen collected and the total neonicotinoid concentration. Given the uncertainty over the contamination of wild plants, it was not clear what long-term chronic neonicotinoid exposure was from pollen or nectar over a whole season. A number of studies have attempted to quantify the levels of neonicotinoids in bee-collected pollen and, through microscopic identification of the constituent pollen grains, to identify the major source of neonicotinoid contamination throughout the season. Most of these studies have used honeybee-collected pollen as the model, as pollen traps are easy to fit to apiaries that can be moved into targeted locations and because individual honeybees display floral constancy the origin of collected pollen pellets can be quickly identified.

Where bees collect a greater proportion of wildflower pollen, neonicotinoid concentrations are lower. Botías et al. (2015) measured neonicotinoid concentrations in pollen during the peak flowering period of oilseed rape and 2 months after this period. During peak flowering, honeybees collected 91.1% of their pollen from wildflowers and 8.9% from oilseed rape, with a total neonicotinoid concentration of 3.09 ng/g. In the later period, 100% of their pollen was collected from wildflowers, with a total neonicotinoid concentration of 0.20 ng/g. Cutler et al. (2014) also sampled honeybee pollen from apiaries adjacent to treated and untreated oilseed rape for a 2-week period in July during peak flowering. Honeybees collected low levels of crop pollen, and higher levels of neonicotinoid contamination were found adjacent to treated fields (9.0% wildflower pollen week 1 to 45.2% week 2, 0.84 ng/g) than untreated fields (15.1% wildflower pollen week 1 to 62.5% week 2, 0.24 ng/g). Long and Krupke (2016) collected data over a longer period of time, from May to September, covering the flowering period of maize, the flowering crop at their study sites. At all sites, a high proportion of pollen was collected from wildflowers. Average neonicotinoid concentrations were lowest at non-agricultural sites (93.9% wildflower pollen, 0.047 ng/g), higher at untreated agricultural sites (95.8% wildflower pollen, 0.078 ng/g) and highest at treated agricultural sites (95.3% wildflower pollen, 0.176 ng/g). Alburaki et al. (2015, 2016) found low levels of neonicotinoids when honeybees collected predominantly wildflower pollen, with none detected in loads of 99% wildflower pollen and average neonicotinoid concentrations of 0.04 ng/g in loads of 93.5% wildflower pollen.

Only two studies are available which measured neonicotinoid concentrations in bumblebee-collected pollen and quantified the proportion of pollen collected from wildflowers. Cutler and Scott-Dupreee (2014) placed out Bombus impatiens nests next to neonicotinoid-treated and untreated maize fields. Bumblebees collected a very low proportion of their pollen from maize, less than 1%, in contrast to honeybees which can collect large quantities of maize pollen during its flowering period (Krupke et al. 2012; Pohorecka et al. 2013, though see Alburaki et al. 2015, 2016; Long and Krupke 2016). Levels of neonicotinoid residues were low, at <0.1 ng/g by untreated fields and 0.4 ng/g by treated fields. In contrast, David et al. (2016) placed out five B. terrestris nests adjacent to treated oilseed rape fields, a crop with pollen attractive to bumblebees. Pollen was sampled from nest stores at the end of June. Bumblebees collected an average of 68.1% wildflower pollen and 31.9% oilseed rape pollen.

Thiamethoxam was found in this pollen at an average concentration of 18 ng/g and thiacloprid at an average concentration of 2.9 ng/g. These levels are much higher than the levels found in honeybee-collected pollen from the same study area in the same year of 3.09 ng/g total neonicotinoids, though a much higher proportion (91.9%) of pollen was collected from wildflowers (Botías et al. 2015). Comparisons are difficult because few other studies have assessed neonicotinoid concentrations in bumblebee-collected pollen with reference to pollen origin. Rolke et al. (2016) placed B. terrestris colonies out next to treated oilseed rape fields and found much lower concentrations of 0.88 ng/g of clothianidin in pollen taken directly from returning bumblebees, but the origin of this pollen is unknown. The concentrations found by David et al. are however lower than the levels reported by Pohorecka et al. (2013) and within a factor of 2 of the levels reported by Rundlöf et al. (2015) who found neonicotinoid concentrations of 27.0 and 13.9 ng/g in honeybee-collected pollen respectively, samples which also contained a high proportion of crop pollen.

Overall, these studies show that the highest acute exposure (0.84–27.0 ng/g) comes during the flowering period of insect-attractive neonicotinoid-treated flowering crops in situations where over a quarter of total pollen intake comes from crop plants. Reported values vary by up to two orders of magnitude depending on crop type, date of sample collection, initial strength of neonicotinoid seed coating and the proportion of wildflower pollen collected. Because only one study has explicitly measured neonicotinoid concentrations in wildflower pollen, it is difficult to judge whether wildflower pollen consistently contains higher or lower concentrations of neonicotinoids than crop pollen. However, when looking at honeybee pollen diets in neonicotinoid-treated agricultural areas outside of the main flowering period of attractive crops, or where flowering crops are unattractive to a specific bee species, neonicotinoid concentrations are generally low, in the region of 0.04–0.40 ng/g from pollen diets composed of 95.3–100% wildflower pollen (Cutler and Scott-Dupreee 2014; Botías et al. 2015; Long and Krupke 2016; Alburaki et al. 2016). Whilst the highest levels of acute exposure come from pollen diets containing a proportion of crop pollen, because honeybees collect pollen over the whole season, total exposure to neonicotinoids may primarily be determined by concentrations in wildflowers. Botías et al. (2015) calculated, based on pollen collected in June and August, that 97% of the total neonicotinoids present in pollen were of wildflower origin. Non-crop plants surrounding agricultural areas represent an additional and chronic source of neonicotinoid exposure.

The risk of neonicotinoid exposure from succeeding crops was identified as a key knowledge gap by the EFSA reports. The available studies suggested that residues in succeeding crops are below LOQ, but the data set was limited. Since 2013, few studies have explicitly looked at neonicotinoid levels in untreated crops grown in soil that had previously been used to grow neonicotinoid-treated crops, as most crops will be sown with a new dose of neonicotinoids each year. However, where specific neonicotinoid formulations are changed, this analysis is possible. Botías et al. (2015, 2016) analysed neonicotinoid concentrations in oilseed rape treated with thiamethoxam. The fields had been used to grow clothianidin-treated cereals over at least the previous 2 years. Imidacloprid had not been used for the previous 3 years. Oilseed rape pollen and foliage were found to contain 3.15 and 1.04 ng/g of thiamethoxam, 1.90 and 2.91 ng/g of clothianidin and 0 and 0.23 ng/g of imidacloprid respectively. As clothianidin can be produced as a metabolite of thiamethoxam, it is not possible to comment on the origin of these detected residues. Imidacloprid was absent from the pollen samples, reflecting the time since the last known agricultural use. Given that these compounds can persist in soil for multiple years, the level of exposure from succeeding crops will broadly depend on the date since the last application, as well as the other factors determining neonicotinoid persistence in soil (“Persistence of neonicotinoids in soil” section). However, as demonstrated by the presence of imidacloprid in foliage samples, succeeding crops can take up residues of neonicotinoids remaining from applications made at least 2 years previously. Given the presence of neonicotinoids in annual, perennial and woody vegetation surrounding agricultural land (“Risk of exposure from and uptake of neonicotinoids in non-crop plants” section), and the medium-term persistence of neonicotinoids in soil and water (“Persistence of neonicotinoids in water and transport mechanisms for contamination of aquatic systems” and “Levels of neonicotinoid contamination found in waterbodies” sections), the risk of exposure from succeeding crops is likely to be in line with levels reported from general vegetation in agricultural environments. However, more explicit investigation in this area is required.

Almost all of the studies conducted on the toxicity of neonicotinoids to bees have been conducted on honeybees, Apis mellifera. Fourteen studies conducted up to 2010 were reviewed in a meta-analysis by Cresswell (2011) who concluded that for acute oral toxicity imidacloprid has a 48-h LD₅₀ = 4.5 ng/bee. The EFSA studies (2013a, b, c) reviewed existing studies for acute oral toxicity up to 2013, including both peer-reviewed studies and also private studies that are not in the public domain (summarised in Godfray et al. 2014). These analyses produced LD₅₀s of 3.7 ng/bee for imidacloprid, 3.8 ng/bee for clothianidin and 5.0 ng/bee for thiamethoxam. Equivalent LD₅₀s for acute contact have also been calculated by EFSA (2013a, b, c) for honeybees to be 81 ng/bee for imidacloprid, 44 ng/bee for clothianidin and 24 ng/bee for thiamethoxam.

However, the EFSA reports highlighted a knowledge gap for the effects of neonicotinoids on bees other than honeybees. Arena and Sgolastra (2014) conducted a meta-analysis comparing the sensitivity of bees to pesticides relative to the sensitivity of honeybees. This analysis combined data from 47 studies covering 53 pesticides from six chemical families with a total of 150 case studies covering 18 bee species (plus A. mellifera). Arena and Sgolastra calculated a sensitivity ratio R between the lethal dose for species a (A. mellifera) and for species s (other than A. mellifera), R = LD₅₀a/LD₅₀s. A ratio of over 1 indicates that the other bee species is more sensitive to the selected pesticides than A. mellifera and vice versa. There was high variability in relative sensitivity ranging from 0.001 to 2085.7, but across all pesticides a median sensitivity of 0.57 was calculated, suggesting that A. mellifera was generally more sensitive to pesticides than other bee species. In the vast majority of cases (95%), the sensitivity ratio was below 10.

Combining data for all neonicotinoids (acetamiprid, imidacloprid, thiacloprid and thiamethoxam) and for both acute contact and acute oral toxicity, nine studies covering nine bee species (plus A. mellifera) were found. These studies showed a median sensitivity ratio of 1.045 which is the highest median value of all the analysed pesticide chemical families. The most relatively toxic neonicotinoids to other bees were the cyano-substituted neonicotinoids acetamiprid and thiacloprid as these exhibit lower toxicity to honeybees than the nitro-substituted neonicotinoids imidacloprid and thiamethoxam.

Selecting pesticides covered by the moratorium (excluding acetamiprid and thiacloprid and including fipronil) and including both acute contact and acute oral toxicity, 12 studies covering 10 bee species (plus A. mellifera) were found. These studies showed a median sensitivity ratio of 0.957 which is close to the calculated sensitivity ratio for all neonicotinoids. The greatest discrepancy between honeybees and other bees was found for stingless bees (Apidae: Meliponini). The effect of acute contact of fipronil on Scaptotrigona postica (24-fold greater), of acute contact of fipronil on Melipona scutellaris (14-fold greater) and of acute contact of Thiacloprid on Nannotrigona perilampoides (2086-fold) were the only three cases with a sensitivity ratio of over 10. Stingless bees are predominantly equatorial with the greatest diversity found in the neotropics. No species are found in Europe (Nieto et al. 2014). In contrast, studies on B. terrestris consistently report a lower sensitivity ratio between 0.005 and 0.914, median 0.264. B. terrestris is widespread in Europe and is the most commonly used non-Apis model system for assessing the effects of neonicotinoids on wild bees (see the “Sublethal effects of neonicotinoids on wild bees” section). Differences in bee body weight have been proposed to explain these differences, with sensitivity to pesticides inversely correlated with body size (Devillers et al. 2003). However, this has not been consistently demonstrated and other mechanisms have been suggested such as species level adaptation to feeding on alkaloid-rich nectar (Cresswell et al. 2012) and differential abilities to clear neonicotinoid residues from their bodies (Cresswell et al. 2014). With the limited data available, Arena and Sgolastra could not comment on the strength of these claims.

Spurgeon et al. (2016) calculated various toxicity measures of clothianidin on honeybees, the bumblebee species B. terrestris and the solitary bee species O. bicornis. Acute oral toxicity 48-, 96- and 240-h LD₅₀s for honeybees were 14.6, 15.4 and 11.7 ng/bee respectively. For B. terrestris, the corresponding values were 26.6, 35 and 57.4 ng/bee respectively. For O. bicornis, the corresponding values were 8.4, 12.4 and 28.0 ng/bee respectively. These findings are generally in line with the findings of Arena and Sgolastra, with B. terrestris less sensitive than A. mellifera at all time points and O. bicornis less sensitive at 240 h.

Sgolastra et al. (2016) calculated relative sensitivity to clothianidin to these same three species over a range of time periods from 24 to 96 h. The highest LD₅₀ values were obtained after 24 h for A. mellifera and B. terrestris and after 72 h for O. bicornis. At these time points, O. bicornis was the most sensitive of the three species, with LD₅₀ measurements of 1.17 ng/bee and 9.47 ng/g, compared to 1.68 ng/bee and 19.08 ng/g for A. mellifera and 3.12 ng/bee and 11.90 ng/g for B. terrestris. These results are in line with the values calculated by Spurgeon et al. (except for the 240-h values), with decreasing sensitivity in the order of O. bicornis > A. mellifera > B. terrestris. Together, these studies support the position that small-bodied species show greater sensitivity to neonicotinoids.

Around 2000 bee species are known from Europe. The biology, behaviour and ecology of each of these species differ from those of honeybees. Consequently, extrapolating from the limited toxicological data available for 19 bee species to the effects of neonicotinoids on the wider European fauna is fraught with difficulties given the wide variation in relative sensitivity. Current data suggests that wild bees are equally to slightly less sensitive to neonicotinoids compared to honeybees when considering direct mortality. However, care must be taken when considering individual bee species, genera and families, as different taxonomic groups may show consistently different individual-level sensitivity. Most European wild bees are smaller than honeybees, and there is the potential for them to be more sensitive on a nanogram per bee basis. In general, continuing to use honeybee neonicotinoid sensitivity metrics is likely to be a reasonable proxy measure for the direct sensitivity of the wild bee community to neonicotinoids (Arena and Sgolastra 2014).

In 2013, a number of studies looking at sublethal effects of neonicotinoids were available, predominantly using honeybees as a model organism in laboratory conditions. Blacquière et al. (2012) reviewed studies on neonicotinoid side effects on bees published between 1995 and 2011 with a specific focus on sublethal effects. The authors found that whilst many laboratory studies described lethal and sublethal effects of neonicotinoids on the foraging behaviour and learning and memory abilities of bees, no effects were observed in field studies at field-realistic dosages. Two major studies that substantially contributed towards the initiation and subsequent implementation of the EU neonicotinoid moratorium were published after this review in 2012.

Henry et al. (2012) gave honeybee workers an acute dose of 1.34 ng of thiamethoxam in a 20 μL sucrose solution, equivalent to 27% of the LD₅₀ (see the “Direct lethality of neonicotinoids to adult wild bees” section), then released them 1 km away from their nests and measured their return rate. Dosed bees were significantly less likely to return to the nest than control bees. Whitehorn et al. (2012) exposed B. terrestris colonies to two levels of neonicotinoid-treated pollen (6 and 12 ng/g plus control) and nectar (0.7 and 1.4 ng/g plus control) in the laboratory for 2 weeks before moving them outdoors to forage independently for 6 weeks, aiming to mimic a pulse exposure that would be expected for bees foraging on neonicotinoid-treated oilseed rape. Bees in the two neonicotinoid treatments grew significantly more slowly and had an 85% reduction in the number of new queens produced when compared to control colonies.

Both of these studies have been criticised for using neonicotinoid concentrations greater than those wild bees are likely to be exposed to in the field (see Godfray et al. 2014; Carreck and Ratnieks 2014). The 1.34 ng of thiamethoxam in a 20 μL sucrose solution used by Henry et al. is a concentration of 67 ng/g. Taking maximum estimated concentrations of thiamethoxam in oilseed rape nectar of 2.72 ng/g (see the “Risk of exposure from pollen and nectar of treated flowering crops” section), a honeybee would have to consume 0.49 g of nectar to receive this dose. Honeybees typically carry 25–40 mg of nectar per foraging trip, equivalent to 0.025–0.040 g, some 10% of the volume necessary to receive a dose as high as the one used by Henry et al. Moreover, as honeybee workers regurgitate this nectar at the hive, the total dose consumed is likely to be a fraction of the total amount carried. Consequently, it is extremely unlikely that the findings of Henry et al. are representative of a real-world situation.

The pollen and nectar concentrations used by Whitehorn et al. are much closer to field-realistic levels with the lower treatment within maximum estimated concentrations of imidacloprid in oilseed rape pollen and nectar (see the “Risk of exposure from pollen and nectar of treated flowering crops” section). However, the experimental setup, where bumblebees had no choice but to consume treated pollen and nectar, has been criticised as unrealistic, as in the real-world alternative, uncontaminated forage sources would be available. Studies that have measured residues in both crop and wildflower pollen and have assessed the origin of bee-collected pollen (see “Risk of exposure from and uptake of neonicotinoids in non-crop plants” section) have recorded neonicotinoid concentrations of between 0.84 and 27.0 ng/g in wild bee-collected pollen where a substantial proportion of this pollen is collected from crop plants during their period of peak flowering. Pollen extracted from bumblebee nests contained neonicotinoid concentrations of 6.5 ng/g in urban areas and 21.2 ng/g in rural areas during the peak flowering period of oilseed rape, though the number of nests sampled (three and five) were low. However, other studies measuring levels in pollen taken directly from bumblebees found concentrations of <1 ng/g, so there is still a lack of clarity surrounding true levels of neonicotinoid exposure for wild bumblebees. On the basis of these described concentrations, the results of Whitehorn et al. are likely to be closer to real-world conditions than the findings of Henry et al.

Post April 2013, much work on sublethal effects of neonicotinoids on bees has been carried out on individual honeybees and honeybee colony fitness metrics, such as colony growth, overwintering success and the production of sexuals. This work is beyond the scope of this review, but important recent publications include Pilling et al. (2013), Cutler et al. (2014), Rundlöf et al. (2015) and Divley et al. (2015) who all found limited to negligible impacts of neonicotinoids at the colony level. See also Cresswell (2011) for a meta-analysis of 13 laboratory and semi-field studies conducted before 2011. Various authors note that interpreting the findings of studies on honeybees to wild bees is fraught with difficulty, given the differing size of individual bees and the social behaviour of honeybees that gives rise to colonies containing many thousands of workers.

Nothing was known about the population level effects of neonicotinoids on wild bees in 2013. As a managed domesticated species, population trends are available for honeybees, but no such data are available for wild bees. One study has attempted to investigate the impact of neonicotinoids on wild bee population trends. Woodcock et al. (2016) used an incidence dataset of wild bee presence in 10 × 10 km grid squares across the UK. The dataset is composed of bee sightings by amateur and professional entomologists and is probably the most complete national bee distribution database currently in existence. Sixty-two wild bee species were selected, and their geographic distance and persistence over an 18-year period between 1994 and 2011 was calculated. Neonicotinoid seed-treated oilseed rape was first used in the UK in 2002, and so the authors calculated spatially and temporally explicit information describing the cover of oilseed rape and the area of this crop treated with neonicotinoids. The 62 species were split into two groups—species that foraged on oilseed rape (n = 34) and species that did not (n = 28). Species persistence across this time period was then compared with expected neonicotinoid exposure. Over the 18-year period, wild bee species persistence was significantly negatively correlated with neonicotinoid exposure for both the foraging and non-foraging groups, with the effect size three times larger for the oilseed rape foraging group.

The characterisation of bees as foragers or non-foragers has one major problem. Many species of bees are obligately parasitic on other bees and do not forage for their own pollen. Some parasitic bees were included in the oilseed rape forager category (n = 2), and some in the non-forager category (n = 12) based on observed nectar visits from a previous study. Some of the parasitic bees in the non-forager group are parasitic on bees included in the forager group (n = 10/28). Given that these species are highly dependent on their host’s abundance, this classification does not make ecological sense. A decline due to a decline in their host or because of increased direct mortality cannot be separated, introducing an additional confounding issue into the analysis. In addition, given the presence of neonicotinoids in wild plants adjacent to agricultural areas (“Risk of exposure from and uptake of neonicotinoids in non-crop plants” section), the amount applied to oilseed rape is not necessarily a true measure of actual neonicotinoid exposure for wild bees.

Overall, the study suggests that bee species were more likely to disappear from areas with a high exposure to neonicotinoids as measured by the amounts applied as seed dressings to oilseed rape, and that this trend was more pronounced for species known to forage on oilseed rape.