Permethrin Contamination of Sawgrass Marshes and Potential Risk for the Imperiled Klot's Skipper Butterfly (Euphyes pilatka klotsi)

Nontarget effects from mosquito control operations are possible in habitats adjacent to areas targeted by ultra‐low‐volume (ULV) sprays of permethrin for adult mosquito control. We assessed the risks of permethrin exposure to butterflies, particularly the imperiled Klot's skipper, when exposed to ground‐based ULV sprays. Samples of larval host plant leaves (sawgrass) were collected in June (in mosquito season) and January (outside mosquito season) of 2015 from sawgrass marsh habitats of the National Key Deer Wildlife Refuge (Big Pine Key, FL, USA) and analyzed for permethrin. Permethrin detection was higher in June (detected on 70% of samples) than in January (30%), and concentrations were significantly higher in June (geomean = 2.1 ng/g, median = 2.4) relative to January (0.4 ng/g, median = 0.2). Dietary risk for 4th to 5th‐instar larvae was low based on the measured residues. The AGricultural DISPersal model (Ver. 8.26) was used to estimate permethrin residues on sawgrass following ULV sprays (deposited residues) to estimate immediate postspray risk. Estimated deposited residues (33–543 ng/g) were much higher than measured residues, which leads to a higher risk likelihood for butterfly larvae immediately after ULV sprays. The difference between estimated and measured residues, and between the two risk estimations, reflects uncertainty in risk estimates based on the measured residues. Research on modeling deposited pesticide residues following ground‐based ULV spray is limited. More research on estimating deposited pesticide residues from truck‐mounted ULV sprayers could help reduce uncertainty in the risk predictions for nontarget insects like butterflies. Environ Toxicol Chem 2024;43:267–278. Published 2023. This article is a U.S. Government work and is in the public domain in the USA. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.


INTRODUCTION
The Florida Keys (USA) are in a humid tropical climate (Osland et al., 2021), one with conditions favorable for increased mosquito activity and elevated disease transmission likelihood (Liyanange et al., 2022).Ultra-low-volume (ULV) sprays are frequently used for control of flying mosquitoes because the density of droplets in the aerosol cloud (droplets/ mL) produced by ULV sprayers is high enough to ensure that flying mosquitos are exposed to an efficacious dose (Mount, 1970).Mount (1970), in a review of earlier studies on the relation between droplet size and efficacy against adult mosquitoes, surmised that droplet sizes in aerosols from groundbased ULV sprays should be in the range of 5 to 10 µm for effective mosquito control.Based on Stoke's law, the settling velocities for droplets of that size are <0.003m/s.Such settling velocities indicate that the droplets are susceptible to drifting with the wind.In fact, wind is necessary to carry the droplets into areas targeted by the spray, where it can affect mosquitoes.Bonds (2012) estimated that, in the absence of obstructions, droplets with a 10-µm diameter could drift 144 m when the wind speed was 0.44 m/s (~1 mph) and 575 m at a wind speed of 1.78 m/s (~4 mph).Because the label for one product containing permethrin (Aqua-Kontrol 30-30, Veseris) states that applications are permitted at wind speeds up to 4.4 m/s (10 mph), droplets could drift much further than 575 m from the spray source in the absence of any obstructions to air flow.Once applied at a location, the aerosols containing the insecticide drift with the wind into the target area, where it impacts flying adult mosquitoes and could drift further into adjacent areas not targeted by the sprays and thus affect nontarget insects.
The Florida Keys (USA) National Wildlife Refuge Complex (Refuge) permits, as needed, aerial applications of Bacillus thuringiensis israelensis and aerial ULV sprays of the organophosphate insecticide naled (US Fish and Wildlife Service, 2014) for mosquito control.While not permitted on the Refuge, truckmounted ULV sprayers apply the broad-spectrum insecticide permethrin routinely in neighborhoods adjacent to the Refuge.The US Environmental Protection Agency (2007) has classified permethrin as highly toxic to insects exposed either by ingestion or by acute contact and has suggested that sprays of permethrincontaining formulations could reduce beneficial insect populations.Because of concern regarding permethrin's impact on a resident imperiled butterfly species (Bartram's scrub-hairstreak butterfly [Strymon acis bartrami]), the Refuge funded a study to investigate potential permethrin contamination in the Refuge (Pierce, 2011).They found permethrin contamination of leaves in the Refuge ranging from 9 to 730 ng/g dry weight.These data were used in a risk assessment for 4th to 5th-instar butterfly larvae exposed to contaminated foliage and predicted that risk was likely (Hoang & Rand, 2015).Bartram's scrub-hairstreak butterfly is not the only imperiled butterfly species in the Refuge.The Florida leafwing (Anaea troglodyta floridalis) is an endangered species that resided within the Refuge but has not been seen since 2007; currently the only known populations are found in Everglades National Park (southern FL, USA).Other imperiled butterfly species (Miami blue butterfly [Strymon acis bartrami] and Schaus swallowtail butterfly [Heraclides aristodemus ponceanus]) are found elsewhere in the Florida Keys.The risk assessment conducted by Hoang & Rand (2015) utilized permethrin contamination of leaves from uplands in the Refuge, but not of sawgrass in marshes of the Refuge.
Another imperiled butterfly species found within the Refuge is the Klot's skipper (Euphyes pilatka klotsi).A subspecies of the more widespread Palatka skipper (Euphyes pilatka), the Klot's skipper is undergoing evaluation by the US Fish and Wildlife Service (2011) for federal listing as either threatened or endangered.Little is known about the biology of the Klot's skipper.It was first described by Miller et al. (1985), who noted that it was restricted to small patches of sawgrass (its larval host plant) in freshwater wetlands within pine rocklands of Big Pine and Sugarloaf Keys in the Florida Keys.A later report on the Klot's skipper (Florida Natural Areas Inventory, 2023) stated it is endemic to the Florida Keys being known also from Big Torch, Ramrod, and Cudjoe Keys.All the Keys are at least in part within the boundaries of the Refuge.
The adult flight period of the Klot's skipper has not been investigated but may be like that of the parent Palatka skipper.A study on the Klot's skipper in the Refuge focused field efforts in the spring and fall, supposing that adults were most prevalent in these seasons.This agrees with flight periods for populations of the parent Palatka skipper along the Atlantic Coast, but in South Florida three or four broods may be the case for the Palatka skipper, and adults are likely present most of the year (Schweitzer, 2001).This indicates that the Klot's skipper may have three or four broods and that larvae at various developmental stages may be present year-round.No reports describing the length of the egg or pupal stages were found, but the larval stage may last for several weeks (Schweitzer, 2001).
Several limitations were evident in those risk assessments.The first was noted by Hoang & Rand (2015).The risk they estimated did not account for direct pesticide deposition onto the larvae following a spray.The second but not unrelated limitation is that exposure for dietary risk estimates likely did not consider residues immediately after deposition.The greatest dietary risk for larvae would be most likely immediately after a ULV spray when the residues are greatest and have not dissipated.It was not clear that the residue levels used in the risk assessment in Hoang & Rand (2015) represent those immediately after a ULV spray.These limitations point to the uncertainties in the risk assessments that have been conducted for butterflies exposed to pesticides in ULV sprays.
The present study was conducted to determine permethrin contamination of sawgrass in marshes within the Refuge and to assess dietary risk for the Klot's skipper larvae based on that contamination.Because sawgrass is the larval host plant of the Klot's skipper, sawgrass samples were collected and analyzed for permethrin to evaluate a major route of potential exposure for the butterfly.Samples were collected in January and June and analyzed to determine possible relation of permethrin contamination with the periods of the year with lower and higher spray frequency, respectively.The detected concentrations were used to estimate dietary exposure and compared with available toxicity data for larval butterflies as an estimate of risk likelihood.In addition, to provide insight into the uncertainty associated with risk based on the measured permethrin concentrations, risk based on estimated permethrin concentrations immediately after ULV sprays was also determined for a subset of the sampled locations.

Study area
The study took place on Big Pine, Torch, Ramrod, Cudjoe, and Sugarloaf Keys (hereafter referred to as the Keys) in the Florida Keys (Figure 1) because the Klot's skipper is known to inhabit the sawgrass marshes on these Keys.Depending on mosquito abundance, most of the neighborhoods on the Keys are sprayed (truck-based ULV sprays) with Aqua-Kontrol 30-30 at a rate of 2.0 ounces/acre (M.Coss, Florida Keys Mosquito Control District, personal communication, September, 2022).While sprays may occur throughout the year, those sprays occurred most frequently from spring through summer (Figure 2) when mosquito abundance is higher (Hribar, 2019).

Sample collection and permethrin analysis
The greater frequency of sprays in the months from spring through summer indicates the possibility of higher permethrin contamination of sawgrass in marsh habitats adjacent to the neighborhoods during that period relative to other times of the year.To test this hypothesis, sawgrass samples were collected from 57 locations in January during the period of low spray frequency and from 56 locations in June during the period of higher spray frequency (Figure 1).The sampling locations were randomly selected from 101 locations utilized in a study of Klot's skipper biology conducted by Refuge biologists.Because the goal for this sampling design was to determine an overall indication of permethrin contamination in sawgrass marshes, sampling locations were randomly selected to minimize potential bias related to ease of access, proximity to truck spray routes, and knowledge of sprays that may have contributed to permethrin contamination of the location.
One subsection of a sawgrass leaf blade was collected from each of 12 to 17 randomly selected sawgrass plants at each of the sampled locations.The leaf subsection was cut from the plant using precleaned stainless-steel scissors (wiped with isopropanolsoaked wipes between sampling locations) by cutting off the tip of the leaf blade and then cutting again 15 cm below the first cut.Each subsection was 15 cm long and 1 to 2 cm wide.The length of each collected leaf was measured to ensure consistency of leaf surface area among sampled locations, and the number of collected sections varied based on plant availability at the locations.The composited leaf sample mass averaged 4.4 g wet weight (SD = 0.58 g) among all the sampled locations.
Duplicate samples (two samples from a single location) were collected from five locations in each of June and January.The duplicate samples from three of those locations were used to estimate sampling and laboratory analysis precision by comparing the permethrin concentrations measured in the duplicate samples.In June, permethrin concentrations differed by an average of 14.7% (SD = 2.5%) between the duplicate samples.In January, permethrin was detected at only one of the three locations.For that location, permethrin concentrations in the duplicates differed by 28.6%.One of the two duplicate samples from the other two locations were spiked in the field with permethrin to estimate accuracy of the sampling and analysis procedure.Recovery of permethrin from the spiked duplicate samples ranged from 90% to 109%.
All sawgrass samples were packed in a cooler with cold packs and shipped to the US Geological Survey Organic Chemistry Research Laboratory in Sacramento (CA, USA) for permethrin extraction and analysis.Prior to extraction, samples were spiked with 13 C 6 -cis-permethrin as a recovery surrogate.Sawgrass samples were sonicated in approximately 50 mL dichloromethane (enough to cover the samples) for 15 min, two times; the extract was then concentrated to 1 mL under nitrogen.The concentrated extract was loaded onto a graphitized carbon cartridge (Restek CarboPrep 90; 6 mL, 500 mg) and was then eluted with 10 mL of dichloromethane.The extract was evaporated and exchanged into ethyl acetate, reduced to 200, and then 20 µL of internal standard (acenaphthene-d 10 ) was added.Analysis for permethrin was performed by gas chromatography-tandem mass spectrometry (instrument details can be found in Hladik et al., 2016).Recoveries of the surrogate, 13 C 6 -cis-permethrin, ranged from 70.1% to 120.2%, average (±SD) was 89.9% (±15.1).The method detection limit for this analysis procedure was 0.2 ng/g wet weight.All permethrin concentrations on sawgrass are in ng/g wet weight.Raw data for the measured permethrin concentrations and sampling locations are publicly accessible (Bargar, 2022).Permethrin concentrations measured on sawgrass samples collected in January and June were compared by the nonparametric Wilcoxon signed-ranks test (performed in Systat Ver. 13 with a significance level of α = 0.05).

Risk assumptions
Risk for the Klot's skipper butterfly was assessed based on larval consumption of permethrin-contaminated sawgrass.Risk was not assessed for deposition of permethrin onto larvae or adults, or for larvae and adults walking on contaminated surfaces even though those exposure routes are possible.Dietary exposure was estimated under two different scenarios.One was that of larvae feeding on sawgrass contaminated by permethrin at the levels measured on collected sawgrass samples, and the other was for larvae feeding on sawgrass contaminated at levels expected immediately after deposition following a ULV spray.In both scenarios, contamination of the leaves on which the larvae fed was assumed to be uniform.The methods of dietary exposure estimation and of permethrin deposition estimation are described in subsequent sections.
Permethrin toxicity to larval Klot's skipper was estimated based on dietary toxicity values for late 4th-early 5th-instar larvae of three other butterfly species (Hoang & Rand, 2015).Other studies report permethrin toxicity to larvae of additional butterfly species (Giordano et al., 2020;Oberhauser et al., 2006Oberhauser et al., , 2009)), but none were found that report toxicity values on a chemical mass consumed/unit body weight of larvae, which was necessary for comparison with exposure estimates that were in the units of permethrin mass/unit weight of larva.While the toxicity data and resulting risk predictions are most applicable to 4th to 5th-instar larvae, it is possible that earlier larval instars may be exposed.No presumptions for the risk estimation (e.g., safety factor) were made about differences in permethrin toxicity related to species or larval developmental stage.
Risk was evaluated by comparing frequency distributions for the median lethal doses (LD50s) and the lowest-observedadverse-effect doses (LOAEDs) with frequency distributions for the estimated dietary exposures.These comparisons were performed for exposures based on permethrin residues measured on sawgrass samples collected in January and in June and based on permethrin residues estimated to have deposited immediately after a ULV spray.Risk was presumed where dietary exposures exceeded the dietary toxicity values.
Estimated dietary exposure based on measured residues.Larval dietary exposure was estimated using the fol- lowing equation: where R is the permethrin residue measured on a sawgrass sample (ng/g leaf), and CR is the leaf consumption rate for butterfly larvae based on data available in Hoang and Rand (2015).Three different CR values (0.0456, 0.0533, and 0.169 g leaf/g body wt/day) were used to estimate dietary exposure.Dietary exposure was estimated for each location from which sawgrass leaf samples were collected in January and June.
Estimation of deposited permethrin residues.Permethrin deposition was estimated for a subset of the sampled locations for comparison with the permethrin concentrations measured on sawgrass at the respective locations.The AGricultural DIS-Persal model (AGDISP Ver.8.26) was used to estimate deposition for all ULV sprays that potentially contributed permethrin to the locations.A potentially contributing spray was one that likely resulted in permethrin deposition at the sampled location as determined by wind direction (spray location upwind of sampled location), distance between the sampled location and the spray location, and wind speed (adequate to transport aerosols from the spray location to the sampled location).Data on spray date and spray location were provided by the Florida Keys Mosquito Control District (M.Coss, personal communication, December, 2021).Weather data for the spray dates were sourced from Weather Spark (2022).Wind speed, spray droplet settling velocity, and spray nozzle height above the ground (~2 m) were used to determine drift distance for a 10µm droplet (Table 1).Because a 10-µm droplet approximates the average D V0.1 (10% of spray volume is in smaller droplets) for a variety of ULV sprayers and pressures (Hoffman et al., 2009), 90% of the spray volume is in larger droplets that would settle out of the air column more rapidly.Therefore, the estimated drift distance for a 10-µm droplet was presumed to be near maximal for the aerosol and would differentiate between contributing and noncontributing ULV sprays.Parameters entered into AGDISP for ULV sprays on all spray dates are shown in the Supporting Information, Table S1.Meteorological data specific to each ULV spray date (wind speed, temperature, and relative humidity) and needed for estimation of deposition by AGDISP are shown in the Supporting Information, Table S2.Most of the input parameters in that table were the same as those used in the Mickle (2005) study of malathion deposition following ground-based ULV sprays with the exceptions of boom height, droplet size distribution (provided by the Florida Keys Mosquito Control District), spray material pertinent to the present study (Aqua-Kontrol 30-30), weather conditions, specific gravity for the mineral oil carrier, and evaporation rate.It should be noted that the values for some of the model required inputs (smallest droplet diameter, boom pressure, spray flow rate) were outside of the recommended ranges for the model but are characteristic of ground-based ULV sprays.The units for permethrin residue deposition predicted by AG-DISP were µg/cm 2 .
Dietary exposure based on the deposited residues.
Because the modeled deposition data were in the units of µg/cm 2 , they were converted to ng permethrin/gram leaf using the following equation: where leaf MPA is the leaf mass/unit area.The leaf MPA was estimated based on surface area and fresh weight for the collected leaf samples.Leaf surface area was calculated as the product of leaf width and length.As noted previously in the Sample collection and permethrin analysis section, the width of the collected leaf blades was 1 to 2 cm but was not recorded for each collected blade.Therefore, a width of 1.5 cm was used for the calculations.The product of leaf width, length (15 cm), and number of blades in the composite sample was the total surface area for the sample.The calculated total surface area for the composite samples ranged from 270 to 337.5 cm 2 .The mass (g fresh wt) for each sample was divided by the total surface area for the respective sample to estimate the leaf MPA.The average leaf MPA for all collected samples was 0.012 g/cm 2 (SD = 0.0018).
Dietary exposure for larvae assumed to be at the locations was estimated as described in the section Estimated dietary exposure based on measured residues, but using only the highest and lowest of the leaf CRs (0.169 and 0.0456 g leaf/g body wt/day) in Hoang & Rand (2015).The higher rate was for the common buckeye butterfly, and the lower rate was for the white peacock butterfly.

RESULTS
Figure 2 shows the number of ground-based ULV sprays that had occurred on the Keys in each month of 2014 and 2015.Sprays occurred in every month of the 2-year period except in March of 2014 and February of 2015.The greater number of sprays in the spring and summer coincides with greater mosquito abundance in the summer (Hribar, 2019).While fewer sprays occurred in the months prior to the January sampling date relative to the June sampling date, not all the sprays would have contributed permethrin residues to the sampled locations because they may have been downwind of the sampled location on the spray date or the wind speed on the spray date was insufficient to result in spray droplets reaching the sampled location.In the 4 months prior to the January and June sampling dates, 10 sprays potentially contributed permethrin to the locations in January, and 34 sprays potentially contributed permethrin in June.Accordingly, permethrin was less frequently detected at the locations sampled in January (17 of 57) relative to the locations sampled in June (40 of 56).Permethrin concentrations in January (geomean = 0.4 ng/g, median 0.2) ranged up to 16 ng/g for a sample from Ramrod Key whereas concentrations in June (geomean 2.1 ng/g, median 2.4) ranged up to 57.7 ng/g for a sample from Sugarloaf Key.Permethrin was detected (1.7-29.2ng/g) in samples from several locations on Big Pine Key that were not estimated to be within the range of any potentially contributing truck sprays.Those locations were adjacent to a neighborhood with no truck route.Overall, permethrin contamination at the locations sampled in January was less than permethrin contamination at the locations sampled in June (Wilcoxon signed-rank test Z = 3.754, p < 0.001; Figure 3).
Permethrin residues estimated to have deposited onto leaves immediately after a ULV spray were estimated for 18 locations.The number of potentially contributing sprays for each of those location ranged from 1 to 10 (Supporting Information, Table S1).In addition, the distance between the sampled locations and the spray locations ranged from 29 to 1,734 m (Supporting Information, Table S1).Deposited permethrin residues ranged from 0.0004 to 0.0065 µg/cm 2 , which equated to 33 to 543 ng/g wet weight.Figure 4 shows the AGDISP estimates of deposited permethrin residues for all potentially contributing sprays at each location relative to the concentration measured at the respective locations.At most of the locations, the AGDISP estimated concentration exceeded the measured concentration at the respective location by at least an order of magnitude.
Risk for butterfly larvae exposed to the measured permethrin concentrations is indicated in Figure 5.Estimated dietary exposures ranged from 0.023 to 2.70 ng permethrin/g body weight/day based on the residues measured in January and from 0.046 to 9.75 ng/g body weight/day based on the  Risk for butterfly larvae exposed to the deposited residues estimated by AGDISP was greater (Figure 6).The dietary exposure distributions in that figure reflect exposure for larvae with two different leaf consumption rates.None of the dietary exposures based on the lower leaf consumption rate exceeded the lowest of the dietary toxicity values (LOAED of 45 ng/g body wt/ day).Approximately 25% of the dietary exposures based on the higher consumption rate exceeded the lowest toxicity value.
None of the dietary exposures based on either consumption rate exceeded the lowest dietary LD50 value (745 ng/g body wt/day).

DISCUSSION
Sawgrass marshes in the Refuge were contaminated by permethrin, indicating drift from areas targeted by the ULV sprays into adjacent habitats not targeted by the sprays.The highest measured concentration (57.7 ng/g wet wt) was similar to permethrin concentrations reported in another study conducted on FIGURE 5: Distributions of estimated dietary exposures for butterfly larvae exposed to sawgrass at the locations sampled in January and June in relation to distributions for 24-h lethal dietary toxicity values (lowest-observed-adverse-effect dose [LOAED] and median lethal dose [LD50]) for butterfly 4th-5th-instar larvae of three different butterfly species (Hoang & Rand, 2015).
FIGURE 6: Distributions of estimated dietary exposures for butterfly larvae consuming permethrin-contaminated leaves in relation to distributions of reported dietary toxicity values for larvae (4th-5th-instar) of three different butterfly species (white peacock, atala hairstreak, and common buckeye butterflies; Hoang & Rand, 2015).The two distributions for dietary exposure are based on two different leaf consumption rates (0.0456 and 0.169 g leaf/g body wt/day) reported in Hoang & Rand (2015).LOAED, lowest-observed-adverse-effect dose; LD50, median lethal dose.
Big Pine Key (7-713 ng/g dry wt; Pierce, 2011).Relative to the present study, Pierce (2011) collected leaves from several broadleaved plant species in the pine rockland habitat that comprises much of Big Pine Key.No other comparable studies were found in the open literature for comparison with the present study.Permethrin was more commonly detected in June when there is greater spray frequency relative to January, indicating that elevated contamination is more likely during the time of more intensive mosquito control operations.It is important to note that the detection of permethrin at a few locations that were not clearly downwind of a contributing truck spray highlights the possibility of permethrin contamination from applications not conducted by the local mosquito control agency.Misting systems that spray permethrin for mosquito control on residential properties are commercially available for purchase by homeowners or are used by private contractors (Cilek et al., 2008).While many of the sampled locations were downwind of potentially contributing truck spray routes, many were also within 200 m of private residents that may utilize misting systems.Results of the present study demonstrate that permethrin from ground-based ULV sprays may drift into and contaminate habitats adjacent to areas targeted by the sprays.
The measured permethrin concentrations were less than AGDISP estimates.Generally, measured concentrations were an order of magnitude less, most likely due to dissipation following deposition.The amount of time between sample collection and deposition at the sampled locations was unknown but based on application data from the Florida Keys Mosquito Control District could have been up to 1 year.Few studies reporting on permethrin dissipation from foliage over the time frame of the present study were found.Many shorter term studies reported the decline of permethrin residues on foliage following sprays and estimated half-lives ranging from 2 to 14 days (Antonious, 2021;Heshmati et al., 2020;Knepper et al., 1996;Southwick et al., 1983;Sundraram et al., 1992;Torstensson et al., 1999;Willis et al., 1992Willis et al., , 1994)).A pair of longer term studies reported detectable permethrin residues on plants up to 2 years after a spray (Sundraram et al., 1992;Torstensson et al., 1999).Both showed an initial rapid decline within approximately 1 month of the spray, in agreement with the previously mentioned half-lives.That was followed by a much slower decline leaving detectable residues long after a spray.Dissipation of permethrin residues on sawgrass may follow a similar pattern.However, the applicability of these studies to the present study is uncertain.For example, the Sundraram et al. (1992) study occurred in a climate (Canadian forests) unlike that of the present study (subtropical).Regardless, the measured concentrations represent what remains after dissipation of residues deposited following multiple ULV sprays and are likely less than concentrations immediately after the sprays.
Risk for larvae was not indicated based on the measured residues.This contrasts with the risk conclusions of Hoang & Rand (2015), who also estimated permethrin risk for larvae based on residues measured on leaves collected from Big Pine Key.They predicted butterfly larvae would be at risk from consuming permethrin-contaminated leaves.The difference between the risk predictions may be explained by a couple of factors.First, risk was assessed for permethrin contamination of two different habitats.The present study focused on sawgrass marshes while Hoang & Rand (2015) focused on pine rockland.Most of the truck spray routes in the Keys are within and around residential areas that are primarily within pine rockland habitat, meaning that the exposure data in the Hoang & Rand (2015) study were for locations likely closer to the sprays where contamination may be higher.Second, the amount of time for dissipation of deposited residues likely differed between the two studies.Some of the locations in the Hoang & Rand (2015) study were sampled the day after a spray, meaning that only 1 day of permethrin dissipation had occurred prior to sample collection.By contrast, in the present study, at least 10 days had elapsed between the date of sample collection and the most recent contributing ULV spray.Given the reported halflives for permethrin on leaves discussed in the previous paragraph, the dissipation time difference could have resulted in a large decrease of deposited residues before sample collection in the present study.The combination likely led to higher measured residues and greater risk estimates in Hoang & Rand (2015).The only other study we found that evaluated risk to butterflies from deposited permethrin residues reported mortality of monarch butterfly larvae placed onto milkweed plants exposed to permethrin from barrier treatments (Oberhauser et al., 2006).The residue levels associated with those effects were not reported, which does not enable comparison with the residue levels measured in the present study.To our knowledge, no other published studies have estimated dietary risk to butterfly larvae exposed to permethrin-contaminated foliage following a ground-based ULV spray.Based solely on the measured resides, permethrin contamination of sawgrass marshes presents little risk for butterfly larvae.
The risk indicated by the measured permethrin residues was likely lower than the risk immediately following a spray.We attempted to evaluate risk based on deposited residues for a subset of the sampled locations to determine how much the risk based on measured residues might be under-represented.We utilized AGDISP for this purpose, which is used to model deposition following aerial ULV sprays given meteorological, mechanical, and droplet size spectra information for the sprays.Studies of its use to estimate deposition following groundbased ULV sprays are limited.While not initially designed to model deposition following truck-based ULV sprays, there is a suggestion that it is capable of modeling drift and deposition of droplet sizes in the range applicable to truck-mounted ULV sprayers (Teske et al., 2009).However, only one study was found that evaluated AGDISP predictions of malathion deposition following ULV sprays: Mickle (2005) found that AGDISP provided reasonable estimations of deposited residues from aerial ULV sprays, for which the model has been studied (Teske & Thistle, 2004;Teske et al., 2000Teske et al., , 2003)).However, estimations for ground-based ULV sprays were inconsistent and varied with wind speed.When wind speed was >11 kph (3.05 m/s), AGDISP reasonably agreed with measured deposition within 150 m of the spray truck but exceeded deposition at distances 150 to 500 m from the truck.When wind speed was <6 kph (1.67 m/s), AGDISP overestimated deposition at distances <150 m from the truck and agreed with deposition levels 150 to 500 m from the truck.Mickle (2005) presented no data for truck-mounted ULV sprays (modeled or measured) beyond 500 m from the truck, or when wind speed was between 1.67 and 3.05 m/s.In the present study, of the 67 sprays potentially contributing permethrin to the 18 locations, the wind speed was >3.05 m/s for 28 sprays and was <1.67 m/s for 21 sprays.Wind speed for 18 of the 67 sprays was between 1.67 and 3.05 m/s, and 19 of the 67 sprays occurred at distances >500 m from the sampled location, both situations for which Mickle (2005) presented no data.Unless vegetation samples are collected the day of or after a ULV spray, modeling will be needed to estimate the risk of deposited residues for butterfly larvae.More research would help with evaluation of the accuracy of AGDISP predictions of deposition from ground-based ULV sprays at wind speeds permitted for the sprays and up to distances from the sprayer that levels hazardous to nontarget organisms may drift.
Dietary risk was influenced by food consumption rate, indicating its significance for risk predictions.The three rates used in our study were lower than the weight-normalized leaf consumption rates reported for larvae of the cabbage butterfly (Pieris brassicae; Hassan et al., 2023).Those rates, which ranged from 0.3 to 3.28 and 0.43 to 1.4 g/g body weight/day for 4th-and 5th-instar larvae, respectively, varied depending on the kale genotype on which the larvae fed.Finke & Scriber (1988) reported weight-normalized consumption rates of 1.09 to 2.41 mg/mg body weight/day for 4th-instar eastern black swallowtail butterflies (Papilio polyxenes).While no other weight-normalized consumption rates for butterfly larvae were found, the limited amount of information indicates that consumption rate can vary with the species being considered and with diet (e.g., kale genotype).It is notable that the consumption rates used in our study were low relative to the limited number of consumption rates available in the literature, which indicates that the exposure estimates used for the risk predictions in the present study may not be excessive.Studies on leaf consumption rates for skipper butterflies in general and Klot's skipper specifically could increase certainty in the dietary exposure estimates for the Klot's skipper.
The risk prediction in our assessment were for late 4th-to early 5th-instar larvae.Given what is known about biology of the parent Palatka skipper, it is possible that earlier instar larvae in the Refuge would be exposed to permethrin.How might the risk prediction differ for earlier instar larvae?One difference may be related to leaf consumption rate.Consumption rates based on leaf area consumed (Badawi, 1981), percentage of leaf area consumed (Kahn et al., 2019), and leaf mass consumed/ minute (Brown & Raubenheimer, 2003) increased in later developmental instars, being greatest in the final (5th) instar.Less evident is whether consumption rate, normalized for larval mass, also changes with developmental stage.If it does not, then the consumption rates reported in Hoang & Rand (2015) might be applicable to all developmental stages.If consumption rate does change with larval mass, then the reported rates should not be used to infer exposure for earlier instar larvae.This is a significant consideration because weight-normalized consumption rates enable exposure estimation given leaf contamination in units of chemical mass/unit leaf mass.Lavoie & Oberhauser (2004) presented data indicating that weight-normalized consumption rates for monarch butterfly larvae increased from 3rd to 5th instars, but the increase between instars was not statistically evaluated.Hassan et al. (2023) similarly reported weightnormalized consumption rates (categorized as consumption index) were higher for 5th-instar relative to 4th-instar Pieris brassicae larvae, but again the statistical significance of the difference was not evaluated.If weight-normalized consumption rates change significantly among developmental stages, and the changes are consistent among species, then weight-normalized dietary exposures for earlier instar larvae (1st-3rd instars) may be lower than estimated in the present study for 4th to 5th-instar larvae.Dietary risk for earlier instars could be different from that estimated in our study for later instars if weight-normalized dietary exposure is lower for earlier relative to later instars.Further relevant studies would help to determine whether weightnormalized food consumption rates for butterfly larvae vary significantly among developmental stages.
Another potential difference is that the sensitivity of earlier instar larvae to permethrin may be different relative to 4th to 5th-instar larvae.Literature reporting dietary toxicity values with the units of chemical exposure/unit body weight of larvae are limited, and Hoang & Rand (2015) was the only reference we found that reported such toxicity data for permethrin.While weight-normalized toxicity data for earlier instar larvae exposed to permethrin could not be found, the relation between larval instar and toxicity has been reported for other pesticides.Krishnan et al. (2020) reported toxicity (mortality) values for 1st-, 2nd-, 3rd-, and 5th-instar monarch larvae exposed to several insecticides (beta-cyfluthrin, chlorpyrifos, chlorantraniliprole, imidacloprid, and thiamethoxam).Sensitivity was inversely related to developmental stage for some of the insecticides, but not others, and was more evident depending on how residue exposure for the larvae was characterized (i.e., µg/g larva, µg/cm 2 leaf, µg/larva).Klokacar-Smit et al. (2007) also reported an inverse relation that was dependent on the toxicant (Bacillus thuringiensis var.kurstaki, cypermethrin, and Spinosad).These studies indicate that larval sensitivity to permethrin may vary with developmental stage.
The risk predictions we present only considered effects on larval survival.Sublethal effects such as reduced larval growth (Bargar et al., 2020;Oberhauser et al., 2006) as well as increased pupation period and decreased adult size (Kobiela & Snell-Rood, 2020;Oberhauser et al., 2009) have been demonstrated for butterflies after larval exposure.Pertinent to the present study, larvae exposure to permethrin-contaminated milkweed plants resulted in increased developmental times (Oberhauser et al., 2006).However, the residue levels on the leaves in that study were not measured, which does not allow comparison with the exposure levels estimated in the present study.Therefore, interpretation of the risk predictions in our study should consider the possibility of sublethal effects.
The risk predictions in our assessment considered only dietary exposure.Other likely routes of exposure include contact with contaminated surfaces and direct deposition onto the butterfly.Hoang & Rand (2015) noted these factors as reasons their risk prediction might under-represent what is experienced by butterflies following a ULV spray.Estimating risk for butterflies contacting contaminated surfaces would require toxicity data for such an exposure.Such data are available for insects from other taxonomic orders (e.g., Orthoptera and Diptera) but were not found for butterflies.The available toxicity data for contact exposure could be used to estimate risk for butterflies but with considerable uncertainty because of the taxonomic differences.Bargar (2012) suggested a method to estimate butterfly exposure from pesticides in ULV sprays depositing onto adult butterflies using what was termed the exposure metric.The exposure metric is a ratio of the butterfly surface area and its mass.Given a reasonably accurate estimation of deposition from a model such as AGDISP, the resulting depositional exposure for the butterfly would be in the same units as the toxicity data (µg/g butterfly) and would be useable for risk estimations.Butterflies in habitats potentially contaminated by permethrin from ULV sprays would experience, as larvae, exposure from consuming contaminated leaves as well as direct deposition and contact with contaminated surfaces (larvae and adults).The additional exposures would presumably increase risk likelihood relative to dietary exposure alone.
The Klot's skipper butterfly is currently undergoing evaluation for federal listing.The only known populations of this butterfly are limited to sawgrass marshes in the Florida Keys, primarily within the boundaries of the Refuge.The present study found permethrin contamination of the larval host plant sawgrass, indicating a high likelihood for their exposure to an insecticide used for the control of adult mosquitoes.Risk based on the permethrin residues measured on sawgrass was low.Because the measured concentrations may not have represented concentrations on sawgrass immediately after a ULV spray, risk was also estimated based on resides that deposited onto sawgrass following the sprays.Greater dietary risk was likely based on the deposited residues.Additional research is needed related to estimation of droplet deposition from truckbased ULV sprays, permethrin toxicity to butterfly larvae, and leaf consumption rates by butterfly larvae at different developmental stages.Such data would increase certainty in permethrin risk estimations for the Klot's skipper butterfly and other butterfly species potentially exposed following ULV sprays conducted for adult mosquito control.
Supporting Information-The Supporting Information is available on the Wiley Online Library at https://doi.org/10.1002/etc.5783.

FIGURE 1 :
FIGURE 1: Proximity of truck spray routes (black lines) along which ultra-low-volume sprays of permethrin occur to the locations from which sawgrass samples were collected for analysis of permethrin.White and black dots represent locations sampled in January and June, respectively.Locations sampled in both months appear as a black dot surrounded by a white ring.

FIGURE 2 :
FIGURE 2: Total number of permethrin applications that occurred by month on the five Keys (Big Pine, Cudjoe, Ramrod, Sugarloaf, and Torch) in 2014 and 2015.Spray application data were provided by the Florida Keys Mosquito Control District (M.Coss, personal communication, December, 2021).

FIGURE 3 :
FIGURE 3: Box and whisker plots of permethrin concentrations measured on sawgrass samples collected in January and June, 2015.Permethrin concentrations in June were significantly higher than in January (Wilcoxon signed ranks test, α = 0.05).

FIGURE 4 :
FIGURE 4: Permethrin concentrations (mean ± SD) estimated to have deposited onto sawgrass leaves immediately after ultra-low-volume (ULV) sprays (black columns) relative to concentrations measured on leaves collected from the respective location (white columns).The AGricultural DISPerson model (Ver.8.26) was used to estimate deposited residues.White columns with error bars indicate locations at which two sawgrass samples were collected, and black columns with error bars indicate that multiple sprays may have contributed permethrin to the sampled location.