Prior to fortification, the Regina heavy clay soil (10 g/kg sand, 290 g/kg silt, 710 g/kg clay; 31 g/kg organic matter), which was collected from a location remote from herbicide application, was analyzed for the active ingredients and hydrolysis products of interest using an extraction solvent system [25] shown to be effective for the extraction of aged herbicide residues from soil. No active ingredients or hydrolysis products were detected in the soil and this batch of soil was used for all desorption experiments.

The heavy clay soil, which had been finely ground using a mortar and pestle, was fortified as follows: a methanol solution (2 mL; 50 μg/mL of each active ingredient and hydrolysis product) was added to a glass vial containing 10 g of the ground clay soil resulting in a fortification level of 10 μg/g for each chemical. Methanol (3 mL) was then added to the vial and the mixture stirred to create a homogeneous slurry, and the vial placed in a fume hood overnight to permit evaporation of the methanol. The dry fortified soil was then mixed to a fine consistency using a stainless steel spatula and the vial capped and wrapped with aluminium foil. The fortified soil was then maintained at room temperature in the dark for three weeks to “age” the herbicide residues.

The desorption experiments were carried out in the fall (September/October) when it was expected that herbicide concentrations in ambient air would be low, relative to earlier in the growing season. High-volume air samplers (Model PS-1, General Metals Works, Village of Cleves, OH, USA), designed to collect airborne organic chemicals, were used in this study. Each sampler was equipped with a PUF/XAD-2 resin composite cartridge preceded by a glass fibre filter [11]. The air flow through the samplers (12.5 m³/h) was calibrated in the field using a Sierra-Misco Inc. (Berkeley, CA, USA) orifice head and was monitored daily by integral Venturi/Magnehelic assemblies. In each experiment, the samplers were operated continuously for a 7-day period resulting in a total air flow of ∼2100 m³/week. Weekly sampling had been used in a previous study [26] because it permitted seasonal sampling (spring, summer, fall) without overtaxing the capacity of the analytical laboratory. The intake of each sampler was 1 m above ground level. The sampling unit in the high-volume PS-1 samplers consisted of a 102-mm dia. borosilicate microfibre filter (Micro Filtration Systems, Dublin, CA, USA) in front of a PUF/XAD-2 resin composite cartridge. The cartridge contained 25 mL of XAD-2 adsorbent resin (Caledon Laboratories Ltd., Georgetown, ON, USA) between 25-mm and 50-mm PUF plugs housed in a 60-mm i.d. by 125-mm long glass cylinder with a stainless steel wire mesh support screen. The filter and the composite cartridge collected particulate and vapour components of the air, respectively. Prior to assembly of the sampling cartridges for field use, the PUF plugs and the XAD-2 adsorbent resin were solvent (acetone) cleaned and monitored for background interferences.

The herbicides selected for study are extensively used for weed control in crop production on the Canadian prairies. The three preemergence-applied active ingredients (trifluralin, triallate and ethalfluralin), which are applied to the soil surface prior to emergence of the crop, are generally incorporated into the soil to minimize vapour loss due to their relatively high vapour pressures (Table 3). Their dissipation from soil is characterized, in part, by a rapid initial vapour loss followed by further vapour loss over the growing season when soil surface moisture conditions are adequate [8].

With postemergence-applied active ingredients, the proportion of the application not intercepted by the crop/weed canopy and depositing on the soil surface is dependent upon the coverage provided by the crop/weed canopy. In the present study, three amine salts (2,4-D DMA, MCPA DMA and dicamba DMA) and four esters (2,4-D 2-ethylhexyl, diclofop methyl, fenoxaprop ethyl and bromoxynil octanaote) of three carboxylic acids and one phenol (bromoxynil) were investigated. After field application, esters and salts of acid herbicides are subject to relatively rapid hydrolysis to the free acid on soil and plant surfaces and in aquatic ecosystems. For example, the half-lives of 2,4-D iso-octyl ester [7], diclofop methyl ester [28] and bromoynil octanoate [9] on the soil surface were ∼3 days, <1 days, and ∼4 days, respectively (Table 3). The corresponding half-lives on plant (wheat, Triticum aestivum L.) surfaces were also short (∼3 days [7], <2 days [10] and <2 days [9]). Rapid ester hydrolysis is also expected following application droplet drift into aquatic environments. In natural waters, half-lives of ∼2 days have been reported for 2,4-D iso-octyl and butoxyethyl esters [29] and bromoxynil octanoate [30]. The dissociation of the salts of the acid herbicides to the corresponding free carboxylate anion is generally considered to be rapid. Consequently, with the exception of fenoxaprop ethyl ester, the free acids or hydrolysis products of six active ingredients including two additional hydrolysis products (mecoprop and clopyralid) were also investigated. In total, the active ingredients/hydrolysis products studied represented a wide range in vapour pressures, soil sorption coefficients and field half-lives (Table 3).

Recoveries of the active ingredients/hydrolysis products from the fortified PUF/XAD-2 resin composite cartridges were generally higher than those from the glass fibre filter/soil units (Table 4). For example, mean recoveries of the hydrolysis products from the PUF/XAD-2 resin cartridges ranged from 72 to 99% whereas those from the filter/soil unit ranged from 40 to 91%. This trend was also evident for the preemergence- and postemergence-applied active ingredients. With respect to the hydrolysis products, the lowest recoveries were obtained for dicamba and clopyralid, most likely due to lower extraction efficiency because of their lower pK_a values (Table 3). Standard deviation values were generally 15% or less.

Although ten desorption experiments were carried out, the number of times that desorption of each herbicide/hydrolysis product was measured varied from ten to three (Table 3). Both the PUF/XAD-2 resin composite cartridges and the glass fibre filter/soil units were analyzed with the intent of trying to achieve a mass balance; that is, of the approximately 5 μg of each active ingredient/hydrolysis product applied with the fortified soil to the filters, to account for what proportion of that was desorbed and captured on the PUF/XAD-2 resin cartridge and what proportion was retained on the filter/soil unit. However, achieving a mass balance was not realistic because recoveries of the active ingredients/hydrolysis products from the PUF/XAD-2 resin cartridges and the filter/soil units at this fortification level were not quantitative and differed significantly (Table 4). Therefore, the amount of each active ingredient/hydrolysis product captured on the PUF/XAD-2 resin cartridges (corrected for the amount captured from ambient air concentrations) was considered to be the amount desorbed from the filter/soil unit and was expressed as a proportion of the corresponding amounts added with the fortified soil to the filter/soil unit. For most of the active ingredients/hydrolysis products, the standard deviation values associated with the recoveries from the fortified XAD-2 resin cartridges and the filter/soil units generally exceeded the corresponding proportions desorbed (Table 3).

Each of the desorption experiments was designed such that two high-volume air samplers were used as controls to determine herbicide active ingredients or hydrolysis products present in the air at the time the desorption experiments were carried out. Ambient air concentrations of the dimethylamine salts of 2,4-D, dicamba and MCPA were not detected during the desorption control experiments nor were the hydrolysis products 2,4-D, MCPA, mecoprop, dicamba, bromoxynil, diclofop and clopyralid. However, background concentrations of the preemergence-applied and postemergence-applied active ingredients were detected, most likely because of fall application. Generally, the amounts of each active ingredient captured on the PUF/XAD-2 resin composite cartridges of the control samplers were of similar magnitude. The lowest background amounts (a few nanograms) were detected for diclofop methyl ester (detected in 2 of 10 background samples) and fenoxaprop ethyl ester (detected in 1 of 2 background samples). Higher amounts (<100 ng) were detected for trifluralin, ethalfluralin, 2,4-D 2-ethylhexyl ester and bromoxynil octanoate. The maximum amounts (<1000 ng) were detected for triallate. The average of the amounts detected on the PUF/XAD-2 resin composite cartridge in the control samplers was subtracted from the corresponding amounts detected in samplers with the fortified soil.

There was no detectable desorption of the dimethyamine salts of 2,4-D, dicamba and MCPA from the fortified heavy clay soil contained in the glass fibre filter/soil unit during high-volume air sampling over a 7-day period. This was not surprising because the polar dimethylamine salts, which have negligible vapour pressures under field conditions (Table 3), would be strongly sorbed to the clay soil and, as salts, no partitioning to the gas phase during sampling would have occurred [34].

Bromoxynil was the only hydrolysis product desorbed in measurable amounts (∼0.1%) even though its vapour pressure is very low (Table 3). Desorption most likely occurred because it is less polar than the other hydrolysis products. Bromoxynil is a phenol and, compared to the other five hydrolysis products which are carboxylic acids, is a relatively weak acid (see pK_a values, Table 3) and therefore less polar. Even dicamba and clopyralid, whose vapour pressures are approximately three orders of magnitude greater than that of bromoxynil, were not desorbed. Thus, the polarity of 2,4-D, MCPA, mecoprop, dicamba, diclofop and clopyralid appears to have played a greater role in their desorption than their relative vapour pressures.

The three preemergence-applied active ingredients consisted of one thiocarbamate (triallate) and two dinitroaniline (trifluralin and ethalfluralin) compounds. The four postemergence-applied active ingredients (bromoxynil octanoate, diclofop methyl ester, 2,4-D 2-ethylhexyl ester and fenoxaprop ethyl ester) are all esters. All of these active ingredients are relatively non-polar and it is apparent from the data in Table 3 that the proportion of each active ingredient desorbed was dependent on its vapour pressure. Of these seven active ingredients, the three preeemergence-applied active ingredients have the highest vapour pressures and are desorbed in greatest proportion. Approximately 50% of trifluralin triallate and ethalfluralin were desorbed from the heavy clay soil under the conditions of high-volume air sampling. The highest vapour pressures among the four ester active ingredients were those of bromoxynil octanoate and 2,4-D 2-ethylhexyl ester. Their vapour pressures were more than an order of magnitude lower than those of the preemergence-applied active ingredients. They were also lower than the vapour pressures of dicamba and clopyralid which were not desorbed, providing another example of the importance of polarity with respect to desorption. Approximately 5% of bromoxynil octanoate and 2,4-D 2-ethylhexyl ester were desorbed from the clay soil. With a vapour pressure about an order of magnitude lower than those of bromoxynil octanoate and 2,4-D 2-ethylhexyl ester, approximately 2% of diclofop methyl ester was desorbed. Finally, fenoxaprop ethyl ester, whose vapour pressure was about two orders of magnitude lower, was less than 1% desorbed. The desorption experiments were carried out in September/October when average temperatures were 11.9 and −3.9 °C, respectively. Consequently, because of decreased vapour pressures, desorption of the active ingredients during high-volume sampling may have been somewhat lower than what may have occurred under warmer temperatures in late May to early July when herbicides are normally applied in the prairie region of Canada.

The relationship between vapour pressure and the proportion of active ingredient desorbed from the clay soil was investigated further by plotting the Log (percent desorbed) versus the Log (vapour pressure) (Figure 1). The plot shows a strong correlation between the percent desorbed and vapour pressure (R² = 0.95) and that appreciable desorption occurs once vapour pressures exceed 1 mPa. Thus, for non-polar active ingredients whose vapour pressures exceed 1 mPa, the proportion associated with the particulate fraction in the atmosphere, as determined by analysis of particulates captured on the glass fibre filter during high-volume air sampling, may be underestimated. This would result in a corresponding overestimation of the vapour component captured on the PUF/XAD-2 resin composite cartridge.

Only one soil type was investigated in the desorption experiments in the current study. The soil selected for fortification with active ingredients/hydrolysis products was a heavy clay soil (71% clay, 29% silt and 10% sand) with low (3.1%) organic matter content, typical of the area in which the desorption experiments were carried out. Adsorption/desorption studies, utilizing aqueous soil slurries, have shown that soil sorption coefficients for several herbicides were correlated with the organic matter content of soils [35,36] and wetland sediment [37]. However, vapours of low molecular weight organic compounds sorb mainly to the mineral component of dry low organic carbon content (∼2%) soils [38]. Thus, if soil with different organic matter and clay, silt and sand content had been used, then the extent of desorption of the various active ingredients (Table 3) may have been different.

There was significant variability (standard deviation) associated with the mean proportions of the various active ingredients and the hydrolysis product bromoxynil desorbed during high-volume air sampling (Table 3). Since temperature affects vapour pressure, one source of this variability could be variation in ambient air temperature which ranged from 1.6 to 21.6 °C and −13 to 1.1 °C in September and October, respectively. Relative humidity also varied during the desorption experiments with daily minimum values varying from 17 to 89% and maximum values from 75 to 100%. Chiou and Shoup [38] have shown that humidity affects adsorption of low molecular weight organic compounds to the greatest extent in low organic carbon content (∼2%) soils. In these soils, sorption of vapours of organic compounds is suppressed by water competition for active sites on the mineral component of the soil. Thus, diurnal and day-to-day changes during the 7-day sampling periods may have also contributed to the variability observed in the mean proportions of active ingredients/hydrolysis products desorbed. Finally, in each experiment, it was assumed that the average amount of any active ingredient detected in the ambient air (control) samples would have been collected on the PUF/XAD-2 resin composite cartridge of the samplers containing the fortified soil. Consequently, the desorbed amount in each experiment was “corrected” by subtracting the amount detected in the control sample. However, Bidleman [23] has pointed out that, during high-volume air sampling, vapours may be adsorbed by particles on the filter. In our experiments, vapours of the active ingredients may have sorbed to some extent on the relatively large amount of particulates (0.5 g soil) present on the filter. If so, “correcting” the desorbed amount by the amount detected in the control samples may have also contributed to the variability observed in the average proportion of each active ingredient desorbed.