Essay: Emission and atmospheric fate and transport of pesticides

CMAQ and Pesticide Emission Model (PEM) were linked to evaluate the atmospheric fate and transport of chlorothalonil. CMAQ was updated to account for potential reaction of vapor phase chlorothalonil with the hydroxyl radical and sorption to secondary aerosol. Simulations were conducted for the 2011 summer agricultural growing season with two use scenarios: 1) fungicide application at the maximum label rate to all crops on which it is registered for use and 2) use estimates based on sales and other data compiled by the US Geological Survey (USGS). Predicted daily average gas phase concentrations were within 0.30- to 72-fold and 0.08- to 4.9-fold of gas phase measurements reported in various studies for simulated scenarios 1 and 2, respectively. Simulations indicated that >99% of the chlorothalonil would remain in the vapor phase following emission. With both use scenarios, the maximum daily average gas phase concentrations were below chlorothalonil human health screening levels.
TEXT
Introduction
The emission and atmospheric fate and transport of pesticides have been widely studied (1-10). Physicochemical properties and environmental conditions determine emission rates and if and when a compound will transform, disperse, and whether there is a potential for long-range transport from areas of use (11). In the atmosphere, pesticide residues have been detected both in the vapor and particulate forms and there is a general lack of understanding of whether the emitted or volatilized pesticide remained in the vapor phase or underwent secondary transformation (12-22). There are a number of well-established approaches to evaluate pesticide volatilization (i.e. emission) and transport potential. This includes field programs, laboratory studies, and/or modeling (23-31). Modeling has the inherent benefit of relatively low cost and offering multi-faceted exploration of the likelihood of a compound to enter the atmosphere via volatilization and subsequent dispersion under varying use and environmental scenarios.
First tier assessments are typically achieved using screening level approximations. This was the approach described in draft guidance detailing volatilization screening tools and methodologies to evaluate acute and chronic exposures from airborne pesticides to bystanders that was published by the US Environmental Protection Agency (USEPA) in 2014 (32). The guidance focused on conventional (non-fumigant) pesticides and supplied preliminary screening analyses for 427 active ingredients which either passed or failed a quantitative screen or required no quantitative screen. Placement in each category was based on estimates of atmospheric concentrations provided by the modified Woodrow equation (33) which relies on a chemical’s physiochemical properties, vapor pressure, soil organic carbon-water partitioning coefficient (Koc), and water solubility for soil applied pesticide; but only vapor pressure for foliar applications. Maximum single application rates were assessed. For chemicals placed in the “failed a quantitative screen” category, further investigation into potential exposure from volatilization of the pesticide was required either through field studies and/or higher tier modeling.
In this study, the pesticide emission model (PEM) (34), a higher tier model, was used to generate an emission inventory for use in a modified version of the USEPA’s Community Multiscale Air Quality (CMAQ) model in an assessment of the potential for volatilization and subsequent transport of the widely used fungicide, chlorothalonil. The coupling of an emission model and a regional dispersion model, such as CMAQ, has precedence. In fact, an earlier version of CMAQ coupled to an emission model was used to evaluate the fate and transport of the herbicide atrazine in the atmosphere in the Midwestern USA (35). The CMAQ model has also been used to evaluate the atmospheric fate and transport of PCBs and chlorinated dioxins and furans (36-38).
As one of the most frequently used fungicides, chlorothalonil represented 24% of the total fungicide active ingredient pounds used in the USA in 2008 (40). The compound’s estimated agricultural use in 2011, the year that was the focus of the current study, was just over 10 million pounds across the USA (39). Primary uses of chlorothalonil are on peanuts, potatoes, tomatoes, cucurbits, fruit crops, golf courses, and lawns. Although labeled for use on field crops, like soybean, corn, wheat, and cotton, there is very limited use (39).
Using the linked PEM-CMAQ models, chlorothalonil atmospheric concentrations were predicted at the county-level during the 2011 5-month summer growing season (May – September) across the continental USA using two use scenarios. The first used fungicide application at the maximum label rate to all crops on which it is registered for use. The second used estimates based on county level sales and other data compiled by the US Geological Survey (39). Predicted concentrations were compared to measured ground-level atmospheric concentrations near treated fields from selected field studies in three states, California, Florida, and Minnesota. Daily average concentrations were also compared to chlorothalonil health screening and long-term exposure limits to determine if there were areas that exceeded these thresholds during periods of use, i.e. growing seasons.
Materials and Methods
Models
The PEM used in this study is based on emission models described by Scholtz et al. (41) and Lichiheb et al. (42). The model utilizes dynamic transport processes, fate mechanisms, physicochemical properties and environmental conditions (e.g. temperature, wind speed, solar radiation, precipitation, land cover, leaf area index (LAI), etc.) to determine mass transfer rates from sprayed surfaces to the atmosphere. Pesticide use rates and crop specific properties coupled with meteorology (e.g. temperature, wind speed, etc.) determine the magnitude of pesticide emissions. PEM contains two components – a soil module for pre-emergence applications and a canopy module for post-emergence applications.
Post-emergence chlorothalonil emissions inputs for CMAQ were simulated using the canopy model, which is based on resistances in the canopy boundary layer and the air boundary above the surface (42). Chlorothalonil application rates, number of applications, and application intervals were taken into account for each modeled crop. The concentrations of the pesticide at the canopy boundary layer were determined based on air diffusivity and vapor pressure. The mass transfer coefficient used in calculating the volatilization rate was based on chlorothalonil properties and local meteorological conditions. The initial mass of pesticide was determined from the application rate and the crop interception (specific to crop growth stage). A detailed description of the PEM development is provided in a publication submitted for peer review (34).
PEM uses the same meteorological data that drives the CMAQ atmospheric model, i.e. Weather Research and Forecasting (WRF) output. The WRF gridded data was spatially averaged for each county in the USA to serve as input to the canopy model within PEM, thereby generating county-level emissions estimates. The WRF output was converted into the correct format for CMAQ using the Meteorology-Chemistry Interface Processor (MCIP) version 4.2 (43). This allowed PEM to reproduce the strong diurnal and seasonal emission variations observed for pesticides (41, 44).
The CMAQ model was designed using state-of-the-science capabilities for modeling various air quality issues, including ozone, particulate matter (PM), toxics, acid deposition, and visibility degradation (45-46). The model contains modules representing horizontal and vertical advection, eddy diffusion, gas-phase chemistry, aerosol physics, aqueous-phase reactions and cloud mixing, emissions, and deposition processes. The gas-phase chemical reactions are determined by a modified version of the Carbon Bond IV mechanism (45-47). CMAQ version 5.0.2, which was used in this study, included an improved secondary organic aerosol module (48).
Model Modifications, Linkages and Inputs
The linked CMAQ-PEM modeling system simulated all important processes involving chlorothalonil, including application rates to specific crop fields, volatility and emission into the atmosphere, followed by atmospheric and chemical transport (i.e. advection, wet/dry deposition, chemistry, gas/aerosol partitioning, etc.). CMAQ was modified to track potential for reaction of vapor phase chlorothalonil with the OH radical as well as gas-aerosol partitioning. The CMAQ emission inventory which was provided by PEM has realistic spatial and temporal distributions based on the type and number of fields in each county across the United States and/or documented use.
The dominant atmospheric oxidation path for chlorothalonil is through the reaction with OH radical, as is the case for most pesticides (49). Similar to sesquiterpene and long alkane species, only one reaction product was used to represent the aerosol species (48). Since there is limited information on the oxidation products of chlorothalonil and their yields, the saturation concentration and enthalpy of vaporization of chlorothalonil was used. Additionally, the mass-based stoichiometric yield was set to 1.12 to account for increased mass of the oxidized products (Table 1). In reality, an array of products could be formed from the oxidation of chlorothalonil, which can more appropriately be modeled using the basis set methodology (50-51). However, since minimal information regarding secondary aerosol formation from chlorothalonil is known, only a single aerosol species was used. The parameters used to model chlorothalonil for both gas and aerosol species are provided in Table 1.
Table 1. Chlorothalonil properties used to parameterize the PEM/CMAQ models.
Property/Reaction Value Reference
Molecular weight 266 g mol-1 (52)
Vapor Pressure 5.72×10-7 mm Hg @ 25°C (53)
Henry’s Law Constant 4050 M/atm @ 298.15 K (52)
Diffusivity in Air 0.049 cm2 s-1
Soil adsorption (Koc) 3840 L kg-1 (54)
Leaf photo-degradation rate 0.23 day-1
Leaf penetration rate 0.14 day-1
AI + OH  semi-volatile product k = 6.2 x 10-15 cm3 molecules-1 s-1 (55)
Saturation Concentration
(particle/gas partition coefficient) 4,444.44 µg m=3 (55)
Enthalpy of vaporization 50,000 J mol-1 (55)
Semi-volatile product (gas)  semi-volatile product yield (aerosol) Alpha = 1.12
Three cases were simulated with the linked PEM-CMAQ models. Case 1 modeled only the chlorothalonil vapor phase concentrations based on the emissions from PEM. Case 2 added the oxidation of chlorothalonil by OH. Finally, Case 3 examined the direct partitioning of chlorothalonil from the gas phase to the aerosol phase. In this case, chlorothalonil emissions were emitted directly to the pesticide precursor species (oxidation products from Case 2), which partitioned to the aerosol phase (Figure 1). All three cases were simulated from May 1 to September 30, 2011, using the same emissions inventory and meteorological files.
Chlorothalonil Use Scenarios
Two chlorothalonil use scenarios were used in simulations. In Scenario 1, the PEM model was iteratively processed for chlorothalonil use on each crop and county utilizing application rates, time of treatment and crop growth stages. The crops with labels for chlorothalonil use were selected based on the total amount of acreage harvested in the USA using the 2012 US Agricultural census (56). Table 2 lists the application rates and usage patterns for the selected crops. As described in the USEPA pesticide volatilization screening guidance (32), all rates were the maximum according to crop specific labels. Each state had specific dates for the start of the growing season, i.e. the simulation period. The start times of foliar application were based on the total number of days (TND) before harvest necessary to achieve the intermittent applications and pre-harvest interval for each crop.
(1)
Table 2 also shows the maximum percentage canopy coverage for the selected crops, retrieved from the Pesticide Root Zone Model (PRZM) metadata (57). The canopy coverage, a surrogate for crop interception, was utilized to estimate the fraction of active ingredient applied that was intercepted by the crop canopy. Crop interception was linearly interpolated within the model during each application based on the application window relative to the crop emergence, as shown in Equation 2. PEM compartmentalizes emissions sources and only the fraction of the applied material which was intercepted by the crop canopy was considered to be available for emission. This could lead to under prediction of total volatilization flux if crop canopy interception is less than 100%.
(2)
Typical planting and harvesting dates in each state (58) were used to estimate crop emergence and chlorothalonil foliar application dates. Figure 2 highlights the date of the first chlorothalonil treatment in each state by crop. The states are ordered by the latitude of the geographic centroid, such that southern states are on the top and northern states are on the bottom. The PEM predicted the hourly flux (µg m-2 hr-1) of chlorothalonil into the atmosphere for each modeled crop. The flux was then multiplied by the number of acres of each crop in a given county.
For a given county, the total hourly chlorothalonil emissions from each crop were summed for all modeled crops The emissions were then mapped to the CMAQ grid by evenly distributing them based on the area of each county in a given grid cell. Daily total chlorothalonil emission files were created and combined with the corresponding emission files from the EPA 2011 Version 6 Emissions Modeling Platform (59). Both PEM and CMAQ were simulated using WRF output from the same 5-month simulation period (May 1, 2011 to September 30, 2011).
Daily emission profiles followed a strong diurnal pattern based on changes in temperature. The highest emission rates were in the Midwest where there are a high number of soybean and corn fields. Based on similar application rates, the chlorothalonil fluxes for each crop were similar within each county. The major difference in overall emission rates was due to the total number of acres of each crop in a given county. As noted above although chlorothalonil is labeled for use on crops like soybean and corn, it is rarely used on these crops (39).
Simulations were also completed with Scenario 2, which provided great agro-realism as county level 2011 pesticide use estimates for chlorothalonil from the USGS National Water-Quality Assessment (NAQWA) database (39) was used to weight PEM emissions for each county. PEM total applied chlorothalonil was determined based on the application parameters in Table 2, along with the acres of each crop type in a given county. If there was no use reported in a given county, PEM emissions were weighted by a factor of zero. All other PEM emissions were weighted by the county PEM total applied divided by the total use estimate. There were 109 counties with no corresponding PEM applications/use. The use scenarios for the continental USA are compared in Figure 3.
Results and Discussion
With Scenario 1, i.e. based on maximum label application rates, the average simulation period and maximum daily average gas-phase chlorothalonil concentrations predicted for Case 1 (no OH reaction or aerosol formation) are shown in Figure 4. As mentioned previously, chlorothalonil is not used on 100% of each crop field and it is infrequently used on soybeans, corn, and other field crops. Despite maintaining these highly conservative assumptions, predicted average chlorothalonil concentrations in air for the simulation period (summer) were relatively low. The maximum growing season and daily average concentrations were approximately 16,000 ng m-3 and 800 ng m-3 respectively.
Scenario 1 simulations only slightly changed for Case 2 relative to Case 1 as on average, the maximum decrease in chlorothalonil concentrations across the US due to OH oxidation was only 0.2 ng m-3, a decrease of <1% . This is indicated in Figure 5 which provides the average difference in the predicted gas phase concentrations of chlorothalonil between Cases 1 and 2 and in Figure 6 which shows the predicted average aerosol concentrations for the simulation period. Chlorothalonil’s limited oxidation was due to the low predicted chlorothalonil concentrations and the compound’s low OH reaction rate. It is on the order of 3-4 times slower than other species (e.g. isoprene, toluene, xylene, etc.). Using the aerosol parameters listed in Table 1, a similar amount of aerosol was generated (0.3-0.4 ng m-3) as compared to the amount of chlorothalonil oxidized.
Further analysis examined the amount of chlorothalonil in the vapor phase that could potentially partition directly into the aerosol phase (Case 3) under the assumptions of Scenario 1 use rates. Chlorothalonil emissions were emitted directly into the oxidation reaction products added to the model in Case 2. This allowed all chlorothalonil emissions to be available for partitioning directly into the aerosol phase without the need to first be oxidized by OH. The predicted simulation period average vapor phase and aerosol concentrations are shown in Figure 7. The aerosol was approximately 1000-fold less than the vapor phase concentrations indicating that there was minimal partitioning of chlorothalonil directly to the aerosol phase. Based on the analyses above, vapor phase chlorothalonil reaction (Case 2) or partitioning into the aerosol phase (Case 3) appeared to be very small. Therefore, only Case 1 was reanalyzed using the more realistic emissions generated with use Scenario 2, i.e. chlorothalonil use rates estimated from USGS NAQWA data (39).
Figure 8 shows the simulation period average using Scenario 2 use rates and Figure 9 the monthly averages during the 5-month simulation. Predicted chlorothalonil concentrations were strongly dependent on the application dates for each crop and field. As indicated in Figure 2, the application dates varied based on the growing season for each state and crop. In May, chlorothalonil was only applied to crops in the Southern states, as the growing season had not started in the Northern states. By August and September, the Northern states also had relatively high chlorothalonil emissions and predicted atmospheric concentrations due to the modeling of application to potato fields. Georgia and Florida had the highest predicted concentrations of chlorothalonil in the atmosphere, which is not surprising as they are the two states with the highest use rates (39). For some states (Georgia, Florida, California, etc.), some crop types had application dates prior to the start of the simulation (May 1st). However amounts were relatively small thus impact on the magnitude of simulated maximum daily average during concentrations would be low.
Simulated Compared to Measured Values
The maximum predicted daily-average chlorothalonil concentrations were compared to measured vapor phase concentrations of chlorothalonil near treated crop fields reported in three states. Table 3 lists the maximum measured daily-average concentration for a given location and the predicted maximum daily-average concentration for the corresponding model grid cell. Predicted daily average vapor phase concentrations were within 0.30- to 72-fold (mean= 17-fold) and 0.08- to 4.9-fold (mean=1.0-fold) of vapor phase measurements reported for scenarios 1 and 2, respectively. Results showed that use estimates strongly impacted model outcomes with the more realistic USGS data (39) providing the closest agreement between measured and predicted values.
When comparing the measured to predicted values, it is important to understand the measurement programs: 1) California EPA DPR Air Monitoring Network 2014 Results (1 January – 31 December) (61), 2) Pesticide Action Network North America (PANNA) in Florida (1 October – 6 December 2007) (63), and 3) PANNA in Minnesota (13 June 2006 – 13 August 2009) (64). During the California study, one 24-hr sample was collected each week at three sites. The sites were located at a police station, high school, and airport, potentially far from crop fields (61). Additionally, by only sampling once per week, it was possible that higher concentrations were missed near the times of chlorothalonil application. In the Florida studies, the sampler was located ~20 m to the west and ~20 m to the south (downwind) of fields where Chinese cabbage was cultivated. The test operator observed application of chlorothalonil on several mornings, which corresponded to the highest measured concentrations of chlorothalonil (63). In the Minnesota studies, chlorothalonil was frequently found in samples collected near potato fields, and not detected in samples from a site located far from potatoes, but close to wheat, corn, and soybeans. As the study progressed, samplers were focused on areas near potato-growing operations (64).
Overall, Scenario 2 use estimates provided better agreement with measured results than Scenario 1. Nevertheless, there were in some cases nearly 5- to 10-fold differences. There are several factors that may have contributed to these differences. Chlorothalonil emissions from fields in each county were divided evenly across all grid cells covering the county. Samplers located in large counties, far from treated fields, would be over-predicted by the model since emissions were evenly distributed across the entire county. This is highlighted in Figure 10, where the predicted daily average concentrations were plotted versus measured values (64) at a site in Minnesota where measurements were made close to treated potato fields with the simulated emissions from county-level use estimates distributed across the entire county. Additionally, not all crops were modeled. In the Florida studies, measurements were made near Chinese cabbage fields; however, this crop was not included in simulations. Differences in measured versus simulated concentrations were much larger for Scenario 1, up to 72-fold. Again, in this scenario, applications to soybean and corn fields were assumed in simulations although this is not a common practice. Evaluating Health Risks of Simulated Air Concentrations
To date, there are no established health standards for pesticides in air either at the state or federal level. However, the USEPA proposed a methodology to evaluate potential non-occupational off-target exposure and risk to bystanders due to airborne pesticides (32). In California, the Department of Pesticide Regulation routinely monitor for frequently used pesticides in high use areas within the state and has developed a metric to estimate the potential for adverse health effects by comparing the air concentrations to its health screening levels (65).
The USEPA’s human health risk assessment scoping document indicates that the inhalation toxicological endpoints, defined as the lowest observed adverse effect level (LOAEL) from an acute inhalation study conducted in rats, for chlorothalonil occupational and residential acute, short-term and intermediate-term (1-6 months) exposures are each 2,000 µg m-3 (60). The calculated human equivalent concentrations (HEC), using EPA’s inhalation reference concentration methodology (66), for occupational acute and short/intermediate-term exposure are 1000 µg m-3 and 300 µg m-3, respectively; while the residential acute and short/intermediate-term HEC are 400 µg m-3 and 60 µg m-3, respectively (60). Similarly, the California Department of Pesticide Regulation (CDPR) lists the health screening level for daily and chronic exposure as 34 µg m-3 (61). The time-weighted average occupational exposure limit reported for chlorothalonil is 100 µg m-3 (62).
Since Case 1 combined with Scenario 1 likely generated the most conservative estimates of chlorothalonil air concentrations, it was benchmarked against the endpoints identified above. The predicted values shown in Table 4 highlight that all areas of the country would fall below the exposure limits, even when using the worst-case risk assessment (application to all labeled crops at maximum rates – use scenario 1). This is represented by the quotient of predicted maximum average concentration and toxicological endpoints.