Refining acute human exposure assessment to pesticides in surface water: An integrated data-driven modeling approach.

The substantial spatial and temporal variability of pesticides has led to large uncertainties when determining their peak aqueous concentrations. There is however a lack of large-scale studies dealing with accurate determination of annual maximum daily concentration (AMDC) across the landscape and over time based on the publicly available monitoring data. We developed a novel data-driven approach that firstly used time series modeling to generate AMDCs for qualified water monitoring sites in the conterminous U.S. With feature variables such as pesticide use and land cover compiled into the dataset, machine learning models using eXtreme Gradient Boosting (XGBoost) and Random Forest Regressor (RF) were then developed to estimate AMDCs in surface waters across the U.S. Both models exhibited significant predictability, while a hybrid model consisting of the average predictions by XGBoost and RF model had the highest prediction accuracy (mean absolute error (MAE): 1.23; R2: 0.61). The analysis of permutation variable importance indicated that pesticide use and drainage area were the two most important drivers. Partial dependence analysis revealed that pesticide use, precipitation, cultivated crop land cover and solubility exhibited concentration-promoting effects, whereas drainage area and molecular weight had concentration-demoting effects. Soil adsorption coefficient (Koc) showed nonmonotonic effects. The hybrid model was used to predict and map AMDCs of four example pesticides, including 2,4-dichlorophenoxyacetic acid (2,4-D), atrazine, glyphosate and imidacloprid during 2016-2019 at national scale. The predictive capability was validated using independent monitoring datasets. The fully evaluated approach significantly reduced the uncertainties in modeling annual peak concentrations and served as a valuable solution for conducting geographically oriented, highly refined exposure assessments for pesticides.

Development of a US national-scale, mixed-source, pesticide, rural well database for use in drinking water risk assessment: an atrazine case study.

For pesticide registrations in the USA under the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA), as implemented by the United States Environmental Protection Agency, drinking water risk assessments for groundwater sources are based on standard scenario modeling concentration estimates. The conceptual model for the drinking water protection goals is defined in terms of (1) a rural well in or near a relatively high pesticide use area, a shallow well (4-10 m); (2) long-term, single-station weather data; (3) soils characterized as highly leachable; (4) upper-end or surrogate, worst-case environmental fate parameters; and (5) maximum, annual use rates repeated every year. To date, monitoring data have not been quantitatively incorporated into FIFRA drinking water risk assessment; even though considerable, US national-scale temporal and spatial data for some chemistries exists. Investigations into drinking water monitoring data development have historically focused on single-source efforts that may not represent wide geographies and/or time periods, whereas Safe Drinking Water Act groundwater monitoring data are focused on a community-level scale rather than an individual, shallow, rural well. In the current case study, US national-scale, rural well data for the herbicide atrazine was collected, quality controlled, and combined into a single database from mixed sources (termed the atrazine rural well database) to (1) characterize differences between exposure estimates from standard EPA modeling approaches for specific characterization, (2) evaluate monitoring data toward direct use in US drinking water risk assessments to compliment or supersede standard modeling approaches to define risk, and (3) evaluate monitoring trends a function of time relative to label changes implemented as part of the registration review process. Of the 75,665 drinking water samples collected from groundwater, atrazine was only detected in 3185, a 4% detection rate.

Application of a new approach method (NAM) for inhalation risk assessment.

The US Environmental Protection Agency (USEPA) and other regulatory authorities have been working to utilize in vitro studies with human cells and in silico modelling to reduce the use of vertebrate animals for evaluating chemical risk. Using the Source-to-Outcome framework, a novel mathematical procedure was developed to estimate the human equivalent concentration (HEC) for inhalation risk assessment based upon the relevant aerosol characterization, respiratory dosimetry modelling, and endpoints derived from an in vitro assay using human respiratory epithelial tissue. The procedure used the retained doses at the various areas of the inhalation tract estimated from a computational fluid-particle dynamics (CFPD) model coupled with a simple clearance model. The effect of exposure was derived from an in vitro assay. The magnitude of exposure and the particle size distributions (PSDs) of the external aerosol droplets were obtained from Unit Exposure values published by the USEPA and published monitoring studies, respectively. The Source-to-Outcome approach incorporates external and internal exposure metrics with the toxicity pathway. The information was then integrated to conduct a risk assessment for agricultural operators exposed to products containing chlorothalonil (CTN), a broad-spectrum fungicide. The HECs for three different PSDs considered in this work ranged from 0.043 to 0.112 mg-CTN/L for nasal and oral breathing. These were compared with the estimated average daily exposure concentration for six representative application scenarios. The resulting margins of exposure (MOEs) ranged from 230 to 70,000 depending on the application scenario. This New Assessment Method (NAM) that combined human in silico and human in vitro methods, eliminated the typical uncertainties associated with extrapolation from rodent studies, with their associated interspecies toxicokinetics and toxicodynamics differences. The intraspecies toxicodynamics and toxicokinetics, are still relevant and may need to be used in an inhalation risk assessment. The NAM presented in this work is not chemical-specific and may be applied to conduct an inhalation risk assessment for workers as well as bystanders who could be exposed to aerosol particles of any cytotoxic respiratory irritant.

Co-occurrence of Conazole Fungicide Residues in Raw Agricultural Commodities Sampled by the United States Department of Agriculture Pesticide Data Program.

In this study, the residue data for conazole fungicides were collated and analyzed in all crop samples reported by the United States Department of Agriculture Pesticide Data Program over the period of 2009-2019. Considering all individual samples, the overall detection frequencies (DFs) of conazoles are less than 13%. Among the 18 conazoles, imazalil had the highest overall DF of 6%, followed by tebuconazole and myclobutanil, with 4% each. Conazoles were detected most frequently in raisins with 28% DF, followed by cherries (frozen and fresh) and grapes, with 12, 10, and 10%, respectively. The presence of multiple conazoles in single commodity samples is very low, below 2%. The analyses found no more than four conazoles present in any given sample. Out of the 18 conazoles, 8 of them were not detected in more than 99.9% of the commodity samples from 2009 to 2019 and, therefore, can be eliminated from screening-level cumulative risk assessment for dietary contributions from food items. While conazoles are widely used on food commodities, co-occurrence of conazole residues was observed only in a very limited number of food commodities, including raisins, grapes, cherries (frozen), nectarines, and peaches. Considering the remaining individual food commodities, the co-occurrence of conazole residues in single commodity samples is very low or not even present.

Spray drift deposition comparison of fluorimetry and analytical confirmation techniques.

Tracer dyes are often used as surrogates to characterize pesticide spray drift and it is assumed that they accurately reflect analytical measurement of active ingredients; however, the validity of this assumption remains inconclusive. Consequently, the influence of measurement technique on the magnitude of deposition of spray drift was investigated using spray drift samples evaluated by traditional analytical techniques (HPLC-MS/MS) and fluorimetry (1,3,6,8-pyrene-tetra sulfonic acid tetrasodium salt dye tracer). The experiment was conducted in a low-speed wind tunnel under controlled meteorological conditions. The herbicide mesotrione was sprayed through three spray air induction nozzles (anvil deflector flat fan TTI11004; flat fan AI11004; flat fan AIXR11003). Spray drift deposition samples were collected using stainless steel discs pairs placed side by side in the center of the wind tunnel at distances of 5, 10, 20, 30, and 40 ft (1.5, 3.1, 6.1, 9.1, and 12.2 m) from the spray nozzle. The analytical technique determined pesticide concentration on one disc per pair, and the other was evaluated by fluorimetry. The experimental results, analyzed using the linear split-split plot model, revealed that median deposition concentrations were 15% higher using the tracer dye fluorescence method relative to the analytical method, potentially due in part to procedural recovery inefficiencies of the analytical method (the mean overall procedural recovery result and RSD was 87% ± 6.4% (n = 12). This relationship was consistent and held true for the three nozzle types at all distances within the wind tunnel. © 2021 Society of Chemical Industry.

Statistical Techniques to Analyze Pesticide Data Program Food Residue Observations.

The U.S. EPA conducts dietary-risk assessments to ensure that levels of pesticides on food in the U.S. food supply are safe. Often these assessments utilize conservative residue estimates, maximum residue levels (MRLs), and a high-end estimate derived from registrant-generated field-trial data sets. A more realistic estimate of consumers' pesticide exposure from food may be obtained by utilizing residues from food-monitoring programs, such as the Pesticide Data Program (PDP) of the U.S. Department of Agriculture. A substantial portion of food-residue concentrations in PDP monitoring programs are below the limits of detection (left-censored), which makes the comparison of regulatory-field-trial and PDP residue levels difficult. In this paper, we present a novel adaption of established statistical techniques, the Kaplan-Meier estimator (K-M), the robust regression on ordered statistic (ROS), and the maximum-likelihood estimator (MLE), to quantify the pesticide-residue concentrations in the presence of heavily censored data sets. The examined statistical approaches include the most commonly used parametric and nonparametric methods for handling left-censored data that have been used in the fields of medical and environmental sciences. This work presents a case study in which data of thiamethoxam residue on bell pepper generated from registrant field trials were compared with PDP-monitoring residue values. The results from the statistical techniques were evaluated and compared with commonly used simple substitution methods for the determination of summary statistics. It was found that the maximum-likelihood estimator (MLE) is the most appropriate statistical method to analyze this residue data set. Using the MLE technique, the data analyses showed that the median and mean PDP bell pepper residue levels were approximately 19 and 7 times lower, respectively, than the corresponding statistics of the field-trial residues.