Appropriate sampling methods and statistics can tell apart fraud from pesticide drift in organic farming

Pesticide residues are much lower in organic than in conventional food. The article summarizes the available residue data from the EU and the U.S. organic market. Differences between samples from several sources suggest that organic products are declared conventional, when they have residues—but the origin of the residues is not always investigated. A large number of samples are being tested by organic certifiers, but the sampling methods often do not allow to determine if such residues stem from prohibited pesticide use by organic farmers, from mixing organic with conventional products, from short-range spray-drift from neighbour farms, from the ubiquitous presence of such substances due to long-distance drift, or from other sources of contamination. Eight case studies from different crops and countries are used to demonstrate that sampling at different distances from possible sources of short-distance drift in most cases allows differentiating deliberate pesticide application by the organic farmer from drift. Datasets from 67 banana farms in Ecuador, where aerial fungicide spraying leads to a heavy drift problem, were subjected to statistical analysis. A linear discriminant function including four variables was identified for distinguishing under these conditions application from drift, with an accuracy of 93.3%.

what should be done instead. This is an improvement of the procedure described in Supplementary Fig. 5(c), which was introduced after finding that inspectors were sometimes taking both centre and margin samples from too large areas, leading to results, which were difficult to interpret. 0.041 0.042 0.020 1) Percent of the total number of organic samples taken by the PDP from 2013 to 2019 (3,710 samples) 2) Percent of samples with residues above limit of quantification (LOQ; PDP uses "LOD" = limit of detection, which is identical) 3) Percent of samples with residues above the NOP (National Organic Program) tolerance. The NOP ( §205.671) establishes that products with pesticide residues above 5% of the EPA (Environmental Protection Agency) tolerance (= maximum residue limit) must not be sold with an organic label. The EPA tolerance is different for each pesticide / commodity combination (e.g. 30 mg/kg for Azoxystrobin in potatoes; 5% would be 1.5 mg/kg), therefore the percentage in this column may be low, even when the MCPL is high. When there is no specific EPA tolerance, or the 5% would be below 0.01 mg/kg, 0.01 mg/kg are used as default tolerance (USDA uses ppm, which is identical to mg/kg). 4) Cumulative sum of all residues in all samples of each commodity. 5) Mean cumulative pesticide load per sample (column F divided by column B) 6) Number of samples of domestic (= USA) vs. imported origin. H + I do not always add up to B, because the origin of some samples was not clear. 7) The MCPL for domestic vs. imported was computed only, when at least ten samples of each origin had been tested.

Supplementary
The higher value is highlighted in yellow. Only when the higher value is identified by an asterisk, the difference is significant (based on a one-way ANOVA, with *: p<0.1; **: p<0.05; ***: p<0.01) 8) "Total" refers to all organic samples tested by the program from 2013 to 2019, therefore the values for the commodities listed here do not add up to the totals. 9) fr. = frozen 10) The corrected total MCPL was computed for all organic samples, not only those listed here. Figure 1c for pesticide residues: (mostly) before release to the organic market (Eurofins) and on the (retail and wholesale) organic market (CVUA).

Time of sampling and testing
Since products already on the market are tested later, residue dissipation might explain the lower residue level.
Both datasets refer to fresh fruits and vegetables only. The time span between testing before release to the market and sampling from the market is minimum for these products, and can therefore not explain the lower residue level in the CVUA samples.

Scope of commodities
The definition of "fruits" and "vegetables" in the two databases could be different.
This was indeed the case. Therefore, nuts, mushrooms, herbs, and processed fruits and vegetables were excluded from the Eurofins dataset, because CVUA does not cover these under fruits and vegetables.

Geographic origin of samples
Eurofins could be testing more samples from countries outside the EU This is probably truebut this is exactly part of what is shown in Fig. 1c: when businesses send organic samples from such countries to this laboratory, the purpose is selling the product on the EU (mostly German) market. What is then tested by CVUA, has already undergone the filter process.

Substances covered by multisubstance screening methods
If one laboratory tests for 750 substances, while the other tests for only 400, results of the former can be expected to show higher total cumulated residues.
CVUA says that each sample was tested for 750 substances, while Eurofins tests fresh fruits and vegetables for "approximately 700 substances" (the number may slightly vary from one test to another, depending on the matrix and special customer wishes). If there is any difference because of this reason, the bias should be in favour of CVUAbut this laboratory found extremely low residues in organic produce.

Additional single-substance tests
Inclusion or exclusion of such tests (e.g. glyphosate, dithiocarbamates, ethylene oxide) could bias the results.
Results for glyphosate and dithiocarbamates are included in the "sum of all residues". While CVUA tests all samples for these substances, Eurofins conducts these tests only on demand by customers, meaning that, similar to above, any bias should lead to higher results in the CVUA samples. 2019 was a year when the concern about ethylene oxide (EO) residues in imported products came up in the EU food industry, and many samples were tested for this substance. Since EO residues are often high, this could have biased the overall result. EO was therefore not considered for computing the MCPL.

Non-pesticide contaminants
Such contaminants not originating from agriculture use (see Supplementary Table 4) might be included or not.
These substances were excluded from the cumulated sum of all residues by both laboratories. Since e.g. phosphonic acid is often found at high levels especially in organic fruits, excluding it from both datasets leads to a substantial reduction in total sum of residues.

Number of samples
The number of samples tested by CVUA is relatively small, as compared to Eurofins, especially for organic products. These figures might therefore not be representative.
The number of organic samples from 2019 only, is indeed quite small. However, CVUA is publishing these data every year since 2013. Adding up the samples from these seven years, the laboratory tested 868 organic fruit and 604 organic vegetable samples. The results for both organic and conventional products remained very consistent across these years. This makes the data for 2019 representative.
Supplementary Table 4: Some substances defined as "pesticides" under EU food law, but in most cases not derived from agricultural pesticide use. These substances were not considered in the comparison shown in Figure 1, nor in any other sections of our article.

Substance Explanation
Anthraquinone Used for denaturing seeds to protect them from birds, therefore officially considered a "pesticide". In most cases, however, anthraquinone residues in food come from exposure to smoke during post-harvest handling, or from other sources of air pollution.

Bromide
Bromide is a metabolite of the fumigant methyl bromide, therefore EU food law sets an MRL for bromide. The substance, however, is also found naturally in most plants, therefore the simple presence of bromide in food does not mean it has been fumigated.
Chlorate, perchlorate Several herbicides are chlorate based. Residues chlorate and perchlorate in food, however, are normally derived from drinking water chlorination or from chlorine based disinfectants used for surfaces in the food industry.
Diethyltoluamide (DEET) Insect repellent. Residues are often derived from farm workers using the substances during harvest. Sometimes, sample takers themselves contaminate the samples.
Phosphonic acid Phosphonic acid can be a metabolite of the fungicide fosetyl-Al. It can, however, also stem from phosphonate based fertilisers. In most cases, it turns out impossible to find the origin of phosphonic acid in food. While fosetyl-Al has a very short half-life, phosphonic acid is extremely persistent in soil and plant tissues.

Phthalimide
Phthalimide can be a metabolite of the fungicide folpet. Since folpet quickly degrades to phthalimide, the metabolite is calculated back to folpet. In most cases, however, residues of this substance have to do with packaging or different forms of environmental pollution, not with folpet spraying.   Table 7: Description of 39 variables tested for their usability for differentiating "spraydrift" from "application". The six coloured variables were most promising in the discriminant analysis. After a classification analysis, the four variables in green remained as the best for discriminating between "application" and "drift". Ratio of highest single value in the centre, to the highest value of the same substance in any of the border samples; excluding cases where the maximum in the centre was < 0.03 mg/kg; when the substance was found only in the centre, 0.001 mg/kg was used for the borders because x/0 would not yield a result

5sumcen
Sum of all residues in the centre sample mg/kg

6sumrat
Ratio of (5) to maximum sum of all residues among the samples from the farm Ratio 6a 6sumrat2 Ratio of the sum of all residues in the centre to maximum sum of all residues among the samples from the farm, but excluding cases where the maximum in the centre was < 0.03 mg/kg Ratio 7 7depmi Ratio of the minimum relative deposit of residues, which would be expected at the distance X (using a recognized drift model), and the real relative deposit of residues Ratio 8 to 9 8dep05mi (+ following) As (7), but only for those cases, where the highest single value is higher than 0.05 respectively 0.1 mg/kg Ratio 10 to 15 10depa (+ following) As (7 to 9), but using the average (10 to 12) respectively maximum (13 to 15) of all residues instead of the minimum value

16rmax
Highest ratio of residues in the centre sample, to residues in the different border samples