Neonicotinoids in Water Samples
Imidacloprid, clothianidin, and thiamethoxam were observed seasonally in water samples collected in the Saginaw River watershed. Acetamiprid, dinotefuran, and thiacloprid were not detected in any of the discrete water samples. One or more neonicotinoid was observed at 86% of sites sampled. Imidacloprid was detected at 64% of sites with a maximum concentration of 220 ng/L and median concentration of 27 ± 11 ng/L. Clothianidin was detected at 86% of sites with a maximum concentration of 98 ng/L and median detected concentration of 31 ± 5 ng/L. Thiamethoxam was detected at 71% of sites with a maximum concentration of 32 ng/L and median concentration of 7.9 ± 2 ng/L. The maximum concentrations were observed at Dutch Creek (DC-0010) for all three analytes. Geometric means of neonicotinoid concentrations per site are presented in Table 1.
Imidacloprid was observed in discrete samples in four of seven months sampled, from July-October, with highest detections in July (Fig. 2A). Clothianidin was observed in discrete samples in all months sampled, with maximum concentrations detected in October (Fig. 2B). Thiamethoxam was only observed in June, July, and October sampling events, with the peak in July (Fig. 2C).
Only three sampling locations had adequate USGS water level datasets and sufficient (≥ 2) detections of neonicotinoids during the sampling period. A strong positive relationship between water levels and neonicotinoid concentrations was observed (r2 > 0.75, p ≤ 0.05) for two of these sites (Pine River and Saginaw River at West Center Road); however, data suggests snowmelt in early April may have resulted in a diluting effect wherein this trend was not observed. The third site, Saginaw River at Independence, where water levels were not correlated with neonicotinoid concentration, is located close to Saginaw Bay where river water levels are more strongly influenced by Great Lakes water levels than by upgradient runoff.
Neonicotinoids in Passive Samplers
Imidacloprid, thiamethoxam, and clothianidin were detected at 100% of sites sampled with POCIS. Acetamiprid and dinotefuran were absent in > 95% of samples, with the highest concentrations estimated below the level of detection at 1.7 and 4.6 ng/POCIS, respectively. No detections of thiacloprid were observed. Median mass loading for all sites and deployment times was highest for clothianidin at 20.7 ± 5.2 ng/POCIS, with a range of 3.8-253.3 ng/POCIS. Imidacloprid median mass loading was 20.3 ± 3.0 ng/POCIS, with a range of 2.4–180.0 ng/POCIS. Thiamethoxam detections were lower with a median of 4.7 ± 0.9 ng/POCIS, with a range of 0.9–28.3 ng/POCIS.
POCIS results differed from discrete samples in that detections were found during all sampling periods; although, seasonal trends were similar between both sampling methods (Fig. 2). Peak imidacloprid concentrations were observed from July-August with lower detections in the spring. Clothianidin and thiamethoxam were again observed to be highest in the late summer (September-October) sampling period, and present in low-moderate concentrations for all other sample months.
Linear regressions were used to compare results of passive and discrete sampling methods. Imidacloprid mass accumulation had a strong positive correlation with discrete sample mean concentrations (r2 = 0.86, p < 0.0001) (Fig. 3A). The relationship for July-August was stronger (r2 = 0.95, p < 0.0001) before including the September-October POCIS dataset, possibly due to the extended 49-day October deployment exceeding the recommended 28-day timeframe. A strong positive correlation between clothianidin mass accumulation and discrete sample mean concentrations was also observed (r2 = 0.74, p < 0.001) (Fig. 3B).
Spatial Variability & Assessment
The predominant land cover was cropland, representing 29–83% of land cover at the sample locations. Built (developed) landscapes represented 3.5–27% of land cover types and tree cover (forested) accounted for 5.6–52% of land use coverage. The Dutch Creek drainage area is almost entirely cropland (83%), whereas the Flint River site had nearly equivalent percentages of crop, forested, and urban land coverage. The six Saginaw River sites had nearly identical land coverage estimates. The site with the lowest crop and built area was in the Chippewa River. A summary of site land use coverage is provided in Table 1.
Maximum and geomean mass accumulations (ng/POCIS) were calculated for each sample site and compared to land cover percentages. Crop cover was the best predictor for clothianidin and thiamethoxam concentrations and demonstrated a moderate positive correlation for maximum and mean concentrations (r2 > 0.45, p < 0.01). Total percentage crop and built land covers had a moderate positive correlation with imidacloprid maximum concentrations (r2 > 0.42, p < 0.02). These relationships were observed for both POCIS and discrete datasets.
Of the tributaries sampled, Dutch Creek had significantly higher concentrations of imidacloprid (> 3.5x all sites geomean) and clothianidin (2-4.8x all sites geomean) in both discrete and POCIS during the July 2021 sampling event. In addition to having the highest crop coverage (83%), it also had the lowest percentage of undeveloped area (forested, 5.6%). Another significant source of imidacloprid was the Flint River in July 2021 (1.5-3.5x all sites geomean). In addition to draining a large amount of cropland, the Flint River also drains a large urban landscape. The Bad River was determined to be a source of clothianidin to the watershed (2-4.5x all sites geomean) for most of the growing season. Interestingly, no detections of imidacloprid were observed in the Bad River. Additionally, the clothianidin concentrations found in the Bad River did not have the same seasonal dynamics as the other sample sites and remained elevated in all months except April, suggesting a unique source or fate/transport scenario as compared to the other sample locations. The low detected concentrations of thiamethoxam were not variable enough to distinguish sources but is likely tied to clothianidin as one of its breakdown products. No significant differences in neonicotinoid concentrations were observed in sites upstream and downstream of wastewater treatment plants, although no samples were taken directly from the outfalls.
Aquatic life values comparison
The literature search captured > 100 pertinent articles that were selected for full article review. Articles meeting the stringent data quality requirements outlined in Rule 57 are presented in the Supplemental Material. Several otherwise acceptable studies were rejected due to the use of a formulation or pesticide product with active ingredient as opposed to reagent grade imidacloprid. Water quality standards for imidacloprid, clothianidin, and thiamethoxam are presented (Table 2). From the currently available literature, the most toxic of the neonicotinoids to aquatic life was imidacloprid > clothianidin > > thiamethoxam. The most sensitive organism classes to all three neonicotinoids included mayfly taxa and midges (de Perre et al., 2015; Finnegan et al., 2017; Maloney et al., 2017; Raby et al., 2018a, 2019; Stoughton et al., 2008). Other sensitive organisms to one or more of the neonicotinoids included caddisfly taxa, a species of amphipod (Hyalella azteca), beetles in the genus Gyrinus, and a species of oligochaete *Lumbriculus variegatus) (Lanteigne et al., 2015; Raby et al., 2018a, 2019). All three water quality standards resulted in a tier II derivation which is presented in the Supplemental Material.
Table 2
Water quality standards (ng/L) protective of aquatic life calculated for Michigan, United States, as compared to USEPA aquatic life benchmarks. Standards include final acute values (end of discharge pipe limit), ambient maximum values (24-hour mean limit), and final chronic values (monthly mean limit).
Parameter Name | Final Acute Value (FAV) | Aquatic Maximum Value (AMV) | Final Chronic Value (FCV) | USEPA acute benchmark | USEPA chronic benchmark |
Clothianidin | 580 | 290 | 81 | 11000 | 50 |
Imidacloprid | 720 | 360 | 29 | 380 | 10 |
Thiamethoxam | 1100 | 530 | 280 | 17500 | 740 |
Imidacloprid concentrations fell below the AMV but occasionally exceeded the FCV. Single discrete samples are not sufficient comparisons to an FCV, since it is based on a 30-day average; therefore, evaluation of geometric means for the two highest consecutive sample months were considered (Table 1). In the absence of two consecutive months of detections, geomeans for all sample months were used. This calculation method assumes neonicotinoid concentrations decreased or increased linearly over the month and is supported by passive sampling results showing a strong correlation with discrete sampling. Using discrete geometric means only, a total of seven sites were identified as exceeding the FCV for imidacloprid. No sites exceeded the FCV for clothianidin and thiamethoxam. No sites exceeded the AMV for imidacloprid, clothianidin, or thiamethoxam.
Using the two regressions presented in Fig. 3A, the mass of imidacloprid corresponding with an FCV exceedance is greater than 65–75 ng/POCIS. The sites that exceed both the 29 ng/L discrete FCV and the POCIS mass equivalent of the FCV are the Saginaw River at multiple locations (SG-0030, SG-0040, SG-0059, SG-0052a), Dutch Creek, and the Flint River. One paired sample (for Flint River, September-October) exceeded the POCIS mass equivalent to the FCV (78 ng), but not the discrete concentration (22 ng/L), which indicates the grab sample was likely an underestimate for monthly concentrations. Conversely, one paired discrete sample (Saginaw River, SG-0070) exceeded the FCV in the discrete sample (29.5 ng/L), but not the POCIS mass equivalent to the FCV (63 ng/POCIS). The FCV was not exceeded for any mean clothianidin concentrations (Fig. 3B).