Harvesters in strawberry fields: A literature review of pesticide exposure, an observation of their work activities, and a model for exposure prediction

Abstract

Strawberry harvesters hand-pick fruit that may result in pesticide exposure from hand foliar contact. This paper included a review of publications on harvester pesticide exposure, an observation of their work activities, and development of an alternative model for pesticide exposure prediction. Previous studies monitored the dermal pesticide exposure of strawberry harvesters and found most of the exposure (>70%) was on the hands. Exposure rates (ERs) were calculated as pesticide amount on the skin per hour worked, assuming foliar contact is proportional to daily work hours. Transfer factors (TFs), used for predicting exposure, were calculated by dividing the ER by the amount of dislodgeable foliar pesticide residue. However, the ERs for harvesters working in the same field at the same time varied by as much as 10-fold, and TFs calculated from different studies varied by up to 100-fold. We tested the assumption of foliar contact time being proportional to daily work hours. We observed full work-day activities of 32 strawberry harvesters. We found that their foliar contact time per work minute differed by up to 46%. We suggested using the amount of strawberries picked to predict harvester foliar contact. For all observed harvesters, their foliar contact time per kg of strawberries picked was 35±5 s. This value was similar among harvesters with varying years of experience, of different gender, and using gloves or not. We proposed a predictive model using the amount of strawberries picked to predict harvester pesticide exposure. The exposure predicted by the model is close to the exposure measured in previous monitoring studies (R²: 0.84). The model slope is 0.33±0.03 × 10³ cm²/kg. Model prediction accuracy was confirmed by monitoring captan exposure to harvesters in two fields. The model may be used as a quick screening method to estimate pesticide exposure before conducting complex human monitoring research.

INTRODUCTION

World strawberry production exceeded 4 million tons in 2008, with the United States as the leading producer, accounting for 30% of the production.¹ Almost all US strawberries (92%) are grown in California, where the average yield is 59 tons of fruit per acre (highest in the world), and the average pesticide use is 166 kg (active ingredient) per acre.^{2, 3} As of 2014, 172 pesticide active ingredients in 863 products were registered for use on California strawberries. In 2012, 6373 tons of these active ingredients were used on strawberries, a 16% increase from 2011.³

Pesticide residues are commonly detected on strawberry leaves and fruit. Foliar contact is the primary route of pesticide exposure to strawberry harvesters who may work long hours (e.g., 60 h a week) in the field.^{4, 5, 6, 7, 8, 9} Evaluation of their pesticide exposure is critical to determine the appropriate restricted entry intervals (REIs) into fields after pesticide applications. In past studies, harvester exposure was assessed by either monitoring the amount of pesticide deposited on the skin or by analyzing pesticide metabolites in blood or urine.^{4, 6, 9, 10} These studies, in which different pesticides were monitored and different assessment methods were used, have never been summarized to generate conclusions useful for mitigation strategies.

Previous studies used dermal dosimeters or metabolic biomarkers to monitor harvester pesticide exposure. The harvester exposure rates (ERs) were calculated by dividing the amount of pesticide exposure by the amount of time worked (hours or minutes) per day. This method assumes that the extent of foliar contact was linearly proportional to the amount of time worked. However, the calculated ERs for strawberry harvesters varied by up to 100 times for harvesters working in the same field at the same time.^{4, 6, 10, 11, 12} These variations in the calculated ERs suggest that foliar contact may not be proportional to the amount of time worked. However, there have been no studies that documented strawberry harvester work activities, making it difficult to explain the cause of the variations.

In this study, we observed the work activities of strawberry harvesters. The results of our observations were used to identify factors that may explain the variations of TFs for harvesters from the different studies. We also reviewed available literature that assessed strawberry harvester exposure to pesticides. Based on these past studies, we developed a mathematical model to predict the amount of dermal pesticide exposure and tested the prediction accuracy of the model through an exposure monitoring study.

MATERIALS AND METHODS

Cooperating Farms and Harvesters

The observations were conducted in the spring and summer of 2013 in Monterey and Santa Cruz Counties (Supplementary Table S1). These counties account for almost half of California’s strawberry acreage. A total of 32 harvesters from 5 crews were observed for one full day of work during the peak harvest season.

Harvester Observation

Each day, we observed the work activities of up to six harvesters, focusing on when they picked and packed strawberries. We took four to six video recordings (N=295, average length: 295±93 s) of each harvester picking one flat of strawberries to document the amount of hand contact with the foliage. The video recordings were taken periodically throughout the work day to account for variations in the amount of foliar contact at different times of the day.

Calculation of Foliar Contact Time

For each harvester, total foliar contact time during the day (t_contact, Table 1) was calculated by adding up foliar contact time to pick one flat of strawberries (Eq. (1)),

Table 1 Summary of important equation variables in this paper.

where:

is the time (min) the harvester spent in contact with foliage to pick one flat of strawberries; is the time (min) the harvester spent to pick one flat of strawberries; and % t_contact is the percent of time the harvester’s hands were in contact with strawberry foliage to pick one flat of strawberries, calculated using Eq. (2).

t_one-hand and t_two-hand, that is, the length of time the harvester had one or both hands in contact with strawberry foliage, was determined from video recordings. To account for the difference in the amount of foliar contact between one hand and two hands, we assumed one hand would result in half the contact of two hands.

For the work periods that were not video recorded, we used the video recordings to estimate a harvester’s % t_contact. For flats picked before the first video recording, we used % t_contact from the first video recording; for flats picked between two video recordings, we used the average of % t_contact from two video recordings, one before and one after the picked flat; and for flats picked after the last video recording, we used % t_contact from the last video recording.

To assure accuracy of the results, each video recording was viewed at least eight times (four times for t_one-hand and four times for t_two-hand) to determine the average t_one-hand and t_two-hand. The SDs were within 10% of the average t_one-hand and t_two-hand values. All the video recordings were then watched by another individual who also calculated the average t_one-hand and t_two-hand. The secondary calculations were within ±30% of the initial calculations.

Harvester Monitoring

We monitored harvesters at two fields in Santa Maria, CA, to validate a model developed for pesticide exposure prediction. Both fields were treated with captan 3 days before monitoring, and at each field, six harvesters were monitored for a full day of work (∼9 h).

Pesticides on hands account for most of the total dermal exposure for strawberry harvesters. Participating harvesters wore nitrile gloves (10 mil thickness) when picking strawberries. Their hand exposure was assessed by measuring captan amounts found on the gloves collected at each break (i.e., morning break, lunch break, afternoon break) and the end of the work day. Captan residues on each pair of gloves were retrieved by cutting the gloves and immersing them in 100 ml isopropanol for 30 s with frequent stirring. The isopropanol was then transferred into a HDPE bottle and stored on dry ice before analysis.

Each day, we also collected six strawberry leaf samples from the picked fields using the Precision Leaf Sampler. Each leaf sample contained 40 leaf disks (total leaf surface area: 400 cm²/sample) cut from randomly selected plants in the fields. Each leaf sample was contained in a 4 oz glass jar and stored on ice before analysis.

Analysis of Captan

The isopropanol containing captan from gloves was centrifuged at 2000 r.p.m. for 5 min and an aliquot of the supernatant was injected into an Agilent 6890 gas chromatograph equipped with electron capture detector (GC-ECD).

Dislodgeable foliar residue was removed from the surface of the leaves with Acetonitrile instead of a dioctyl sodium sulfosuccinate (DSS) water solution. During analytical method development, captan was found to quickly degrade at a rate of ∼10% per hour in the DSS water solution. This finding explained a previous study where the amount of degradate tetrahydrophthalimide (THPI) was equal to the amount of captan found in the samples.¹¹ For DFR analysis, each leaf sample containing 40 leaf disks were mixed with 50 ml acetonitrile containing 0.1% acetic acid (ACN/0.1% HAc). The solvent was mixed with leaf disks for 3 min and then decanted. The process was repeated once. The two solutions were combined, centrifuged at 2000 r.p.m. for 5 min, and an aliquot of the supernatant was transferred to a vial for GC-ECD analysis.

Chemical analysis was carried out by introducing the final extracts into the GC inlet at 220 °C. A Zebron ZB-5 capillary column (30 m 145 × 0.25 mm × 0.25 μm; Varian, Sunnyvale, CA, USA) was used for separation and the flow rate of helium was set at 2 ml min. The column temperature was as follows: 200 °C for 1.0 min, ramped to 250 °C at 20 °C per min, and held at 250 °C for 7 min. The elution time of captan was 3.78 min.

Statistical Analysis

Observational data on foliar contact time were interpreted through various statistical methods. First, the normality of foliar contact time for all harvesters were tested using Shapiro–Wilk, Kolmogorov–Smirnov, Cramer–von Mises, and Anderson–Darling analyses with P-value at 0.05. Second, factors that may affect harvester foliar contact time (i.e., work time, work productivity, gender, use of hand protection, and work experience) were evaluated using Best Subset Regression (BSR) and General Linear Model (GLM). Third, the correlation linearity between the foliar contact time and different impact factors was assessed using a generalized additive model.

RESULTS AND DISCUSSION

Review of Literature on Exposure Assessment Studies

In the published strawberry harvester exposure studies, dermal pesticide exposure was estimated by placing dosimeters (e.g., gauze cloth, cotton shirt) on different body parts, and then quantifying the amount of pesticides extracted from these dosimeters.^{4, 6, 9, 10, 11, 12, 13, 14} For hand pesticide exposure, the amount of pesticides deposited on hands in these studies were monitored either by requiring participating workers to wear gloves during the work hours and analyzing pesticides collected on gloves, or by washing the hands of participating workers after the work hours and analyzing pesticides in the hand wash.

These studies found most dermal pesticide exposure (>70%) was by hand contact with strawberry foliage. Krieger and Dinoff⁹ monitored strawberry harvester pesticide exposure only on face and hands, and found 87–98% of the exposure was from hands. Contact to the arms accounts for most of the remaining exposure. Zweig et al.¹⁰ found that the forearms accounted for ~25% of total strawberry harvester dermal exposure to carbaryl.¹⁰ Lanning et al.⁶ found that abamectin levels on harvesters’ long-sleeve shirts were 24.4±10.5% of full-body exposure. Comparing the exposure to the hands and forearms, Zweig et al.¹³ found the amount of captan and benomyl on harvesters’ hands were 5 to 46 and 2 to 32 times the amount detected on the forearms.¹³ Compared with strawberry harvesters, hand exposure accounted for a smaller amount of the total dermal exposure for other agricultural workers in strawberry fields, for example, <5% for pesticide applicators, and for harvesters of other crops, for example, <10% for peach harvesters.^{14, 15} As the amount of pesticide on the hands of strawberry harvesters is much higher than for the rest of the body, accurate monitoring of the duration of hand contact with foliage is crucial for understanding total pesticide exposure.

Pesticide exposure may also be assessed using biomarkers that measure the amount of pesticide absorbed by the human body. Common biomarkers include pesticide metabolites or enzymes in blood or urine.^{16, 17, 18} Urinary levels of THPI, a captan degradate, have been used to assess captan exposure.^{9, 14} Dialkyl phosphates in urine have been used to measure malathion exposure.¹⁶ Using pesticide metabolites as exposure biomarkers assumes that the metabolites are solely from the breakdown of the absorbed pesticide, and ignores the possible formation of these metabolites in the environment. Human uptake of these metabolites from the environment will result in overestimation of pesticide exposure. For instance, strawberry harvesters could be exposed to levels of THPI that may be as high as the parent compound captan.¹⁴ However, in most exposure studies, pesticides and their degradates were never measured simultaneously, and the difference in dermal sorption rates for the pesticides and their degradates was unknown.

In these studies, the amount of pesticide measured by dosimeters or biomarkers was divided by the amount of time worked to determine the work time-based ER (ER_{work time}). The calculation is based on the assumption that harvester foliar contact is proportional to the amount of time worked. Figure 1 summarizes the calculated hand ER_{work time} for strawberry harvesters from different studies where fields were treated with different pesticides, using different application rates, and the harvesters were monitored on different days after application.^{4, 6, 9, 10, 13, 14, 20} Within the DFR range (10⁻³–10¹ μg/cm²) of these studies, the harvesters working in fields with high DFR levels had a tendency to have higher ER_{work time} (Pearson’s r: 0.81). However, for fields with similar DFR levels, the hand ER_{work time} could be one order of magnitude different. In one study, the captan exposure rates on the hands of two harvester crews were 14.32±3.34 and 4.39±2.83 mg/h, even though they were working in fields with similar DFRs (1.4 and 1.7 μg/cm²).¹¹ The ER_{work time} could be up to 20 times different even for harvesters from the same crew.^{10, 13, 14, 19} Krieger and Dinoff⁹ found the amount of captan rinsed from the gloves of harvesters from the same crew ranged from 74 to 1754 μg/day.⁹ There was no explanation given for the variation in the ER_{work time}.

Figure 1

Strawberry harvester hand exposure rates. The rates, expressed as average values and SDs, were plotted against pesticide dislodgeable foliar residues. Data are cited from previous studies.^{4, 6, 9, 10, 12, 14, 20}

To quantify the ease of pesticide transfer from foliage to skin, the ER is divided by the DFR to determine a transfer factor (TF). TF is most commonly used to quickly estimate dermal pesticide exposure based on DFR when human exposure data are not available. TF is expressed as cm²/h or cm²/day for a specific activity on a specific crop, and is considered independent of the pesticides applied. Table 2 summarizes strawberry harvester hand TFs calculated from past studies and shows that the TFs from different studies may vary by as much as two orders of magnitude. Regulatory agencies use different TFs when conducting their exposure assessments. The US Environmental Protection Agency (US EPA) uses a value of 1100 cm²/h to estimate strawberry harvester pesticide exposure, whereas the California Department of Pesticide Regulation (CDPR) uses a value of 1500 cm²/h.^{19, 20}

Table 2 Pesticide transfer factor for strawberry harvesters’ hands.

There are no studies that have measured the foliar contact time of strawberry harvesters. As foliar contact time is time consuming to measure, alternate methods are needed to predict foliar contact time. It is also unclear what factors may affect the amount of foliar contact during work time, making it difficult to explain the variations of ERs and TFs seen in previous studies.

Harvester Work Activity and Foliar Contact

We closely observed 32 harvesters for one full day of work. Their work time ranged from 308 to 565 min, and the amount of strawberries picked ranged from 176 to 400 kg. The recorded foliar contact time of individual harvesters was normally distributed. Detailed information on harvester crews and individual harvesters is provided in the Supplementary Information.

Harvester work time is not always a good predictor of foliar contact time. As shown in Figure 2, harvesters from the same crew may have similar work time, but their foliar contact time differed. In an observation of crew MON-01, the work times of the 5 harvesters were almost identical (277–284 min), but their foliar contact time ranged from 100 to 146 min. For three of the five crews, Pearson’s analysis did not indicate a significant correlation between work time and foliar contact time (Table 3). The regressions of different crews were also different, indicating it is not possible to develop one regression line for predicting the foliar contact of harvesters from different crews.

Figure 2

Correlation of harvester foliar contact time to the work time. Dots with the same shape and color represent harvesters from the same crew and observed on the same day.

Table 3 Parameters from Pearson’s analysis and linear regression.

In contrast, BSR analysis showed work productivity, that is, the amount of strawberries picked, better predicted the foliar contact time (Figure 3). Positive Pearson’s correlations between foliar contact time and the amount of strawberries picked were seen for all crews (Pearson’s r: 0.53–0.91, significance >0.95, Table 3). Among harvesters in the same crew, foliar contact time per kg of strawberries picked showed less variation than foliar contact time per work time. For instance, in crew SCR-01, the foliar contact time per kg of strawberries picked varied by 9% (0.57–0.62 min/kg), whereas the foliar contact time per work time varied by 39% (0.30–0.42 min/min). Different crews had similar regressions of foliar contact time to the amount of strawberries picked. By using work productivity, the predicted foliar contact time was close to the recorded contact time (101±16%, range: 79–141%). For 26 harvesters, the predicted contact time was within ±20% of the recorded value.

Figure 3

Correlation of harvester foliar contact time to the amount of picked strawberries. Plots with the same shape and color represent harvesters from the same crew and observed on the same day.

Statistical analysis showed that using both harvesters’ work time and the amount of strawberries picked gave the best prediction of foliar contact time. Other factors such as the use of gloves, work experience, and gender had minimal impact on harvester foliar contact time (Supplementary Information, Table S2 and Figures S1-S3). Analysis using generalized additive model (GAM) further confirmed that the foliar contact time was linearly correlated to both the work time and the productivity. Based on the above analysis, we developed a model that best quantifies harvester foliar contact (Eq. (3)). Based on this model, the estimated foliar contact time is 85–124% of the recorded time for all harvesters.

where M (kg): the amount of strawberries picked by a harvester and T (min): the length of work time during a day.

Pesticide Exposure Prediction and Validation

Our findings suggest that work productivity is a more accurate variable than work time to predict foliar contact time. Thus, we assume work productivity would also correlate better to harvester pesticide exposure and can be used to calculate a more accurate transfer factor. We tested our assumption first by using the only study that recorded both the work hours and the amount of strawberries picked for individual harvesters.¹² This study supported our assumption. The work productivity of four out of five observed crews correlated better to captan exposure than the work time. The correlation was significantly positive for two crews (Figure 4).

Figure 4

Plots of hand captan exposure amounts against harvester work hours or the number of strawberry flats picked. Harvesters worked in five different fields, and only harvesters with raw data provided are plotted in the figure.

We also summarized all the publications that reported individual harvesters’ ER_{work time} and their picking speed (kg of strawberries picked per hour). With this information, we calculated the ER based on the amount of strawberries picked (ER_productivity) using Eq. (4):

We plotted ER_productivity against the DFRs for the fields where the harvesters worked, and found a positive, linear correlation (Pearson’s r: 0.85, Figure 5). The slope of this linear correlation, that is, TF based on the work productivity, was 1.65±0.15 cm²/flat of strawberries (0.33±0.03 cm²/kg).

Figure 5

Plots of harvesters’ hand pesticide exposure rates against dislodgeable foliar residue. The hand pesticide exposure rate was the average of pesticide amounts divided by flats of strawberries picked.

This linear correlation can be used to predict harvester exposure. We tested the prediction accuracy of our model by monitoring pesticide exposure of harvesters working in two strawberry fields. For both fields, the predicted captan exposure based on work productivity was close to the amount of captan found on the harvesters’ gloves (Table 4). The measured captan amounts on harvesters’ gloves were 6.7–10.6 and 8.0–14.4 mg for the two fields respectively, and this is close to 5.8–11.8 and 11.0–23.9 mg predicted exposure using work productivity. In comparison, using US EPA TF (1100 μg/cm²) underestimated harvester exposure in field 1, and using CDPR TF (1500 μg/cm²) calculated slightly higher exposure than using the amount of strawberries picked in both fields 1 and 2 (Table 4). CDPR TF is higher than US EPA because CDPR used different strawberry studies from US EPA. US EPA TF also included tomato harvesting studies because they assumed tomato harvesting is similar to strawberry harvesting.²⁰

Table 4 Monitored and predicted captan exposure of individual harvester.

This simple model can be used as a quick screening method to estimate the dermal exposure of strawberry harvesters without conducting exposure monitoring. The essential variable of the model, that is, the productivity of each worker (number of flats of strawberries picked), can be obtained from the grower who tracks this information for wage purposes. By analyzing the DFR data, risk assessors can easily estimate pesticide exposure for individual harvesters regardless of the harvester’s gender or work experience. The TF is the same for all pesticides.

Furthermore, combined with pesticide toxicity data, this model can also help regulators determine safe pesticide application rates and REIs after pesticide application. The LC₅₀ through dermal application of captan on rabbits is 4500 mg/kg.²¹ Assuming a safety factor of 10, dissipation half-life on leaves of 9 days, a 24 h REI, harvester’s weight at 75 kg, and his/her productivity at 370 kg strawberries/day, the amount of initial captan deposited on the leaves should be no more than 299 μg/cm².^{22, 23, 24}

This paper is a comprehensive study of strawberry harvester pesticide exposure, including a review of previous work, an evaluation of the results of a strawberry harvester observation study, development of a predictive model of harvester pesticide exposure, and field monitoring of harvesters to validate the model. By reviewing previous literature, we found harvesters working in the same field were exposed to different amounts of pesticides despite similar work time. The differences in exposure may be attributed to differences in the amount of time that harvesters were in contact with the foliage. Our observations showed harvester foliar contact time correlated to the amount of strawberries picked (P<0.01). The correlation fits well for harvesters of both genders, with varying work experience, and using different hand protection, suggesting its utility in using the amount of strawberries picked to predict harvester foliar contact. We propose using the amount of strawberries picked as an alternative to predict harvester pesticide exposure. The prediction model was developed based on previous monitoring studies and validated through a field monitoring study.