1 Introduction

Prediction of pesticide leaching from agricultural lands is complicated. This is because leaching potentials are influenced both by variation in soil structure, by the physical, chemical, and biological parameters of the soil, and by the chemical properties of the specific pesticide as well as timing between precipitation and pesticide application.

Microbial mineralization of pesticides is an important process with respect to pesticide leaching since mineralization is the major process for the complete degradation of pesticides without generation of metabolites. If predictive links between measurements of simple soil characteristics and microbial pesticide mineralization could be established, it would be possible to predict mineralization and consequently identify field-scale areas that may be prone to pesticide leaching (Rodríguez-Cruz et al. 2006; Vinther et al. 2008). In a thorough field-scale study of linuron mineralization, it was shown by multivariate statistics that variation in the initial mineralization rates could be explained by the concerted action of pH, the C/N ratio, and the potassium concentration (Rasmussen et al. 2005). Other studies have indicated that the variability in pesticide mineralization may correlate with soil pH (Bending et al. 2003; Ghafoor et al. 2011), microbial biomass (El Sebai et al. 2007; Walker et al. 2001), and clay and organic carbon content (Vinther et al. 2008).

The large spatial heterogeneity in soil physico-chemical and microbial parameters, however, challenges the establishment and application of such general links. The spatial distribution of potential pesticide mineralization has been documented in many studies at levels ranging from soil microniches to agricultural fields (Badawi et al. 2013; Bending et al. 2006; El Sebai et al. 2007; Gonod et al. 2006; Stenrød et al. 2006; Walker et al. 2001). Most of these studies have investigated pesticide mineralization in the biologically active plough layer. Though variable, the pesticide mineralization in agricultural top soils is often very fast compared to mineralization in deeper layers. This effect has, for instance, been demonstrated for 2-methyl-4-chlorophenoxyacetic acid (MCPA) both at the centimeter scale (Badawi et al. 2013) and at the field scale (Fredslund et al. 2008). For biodegradable pesticides, mineralization, therefore, seems to take place mostly in the plough layer.

In the present study, we have focused on the mineralization potential of the four agricultural herbicides glyphosate [N-(phosphonomethyl)glycine], bromoxyniloctanoate [2,6-dibromo-4-cyanophenyl octanoate], diflufenican [2′,4′-difluoro-2-(α,α,α-trifluoro-m-tolyloxy)-nicotinanilide], and bentazone [3-isopropyl-1H-2,1,3-benzothiadiazin-4(3H)-one 2,2-dioxide] in the topsoil of a loamy, agricultural field. This field has two opposing gradients in clay and organic carbon content, which makes it ideal for testing correlations between soil parameters and pesticide mineralization. This field is part of the Danish Pesticide Leaching Assessment Programme, where continuous field-scale monitoring of pesticide leaching has been carried out since 2000 with the objective to monitor whether pesticides or their degradation products leach to the groundwater under actual field conditions. Therefore, we also wanted to investigate whether the potential for pesticide mineralization in the topsoil could explain the observed differences in pesticide leaching between the northern and southern parts of the field.

2 Material and Methods

2.1 Field Site

The agricultural test field (Silstrup, northwestern Jutland, Denmark) was a conventionally cultivated, loamy field with a cultivated area of 1.69 ha. The climate is coastal, cold temperate. The field has been cultivated as part of a routine agricultural practice with management- and pesticide records dating back to 1983 (Lindhardt et al. 2001; Rosenbom et al. 2015). Glyphosate was sprayed on the field five times since 1983, bentazone was sprayed four times, bromoxyniloctanoate only once, and diflufenican not at all. Application dates for the pesticides and their commercial formulations are shown in Table 1.

Table 1 Pesticide application history. There is no record of which Roundup formulation was applied in 1988 and 1999
Full size table

Two horizontal wells, H1 and H2 (Fig. 1), are located 3.5 m below the surface, and each consists of three screen sections of 18 m. Water samples from the middle screen section of each well (H1.2 and H2.2) have been analyzed for pesticides every month, and the samples from the outer screen sections (H1.1, H1.3, H2.1, and H2.3) have been analyzed twice a year (Rosenbom et al. 2010). During 9 years screening (2000–2009), pesticides were detected in 44 % of the water samples from the middle section of the northern horizontal well (H1) whereas only 5 % of the water samples from the middle screen section of the southern horizontal well (H2) contained detectable pesticide concentrations. In the outer screen sections of the northern well (H1.1 and H1.3), pesticides were detected in 30 and 27 % of the water samples whereas there were no pesticide detections in the outer screen sections of H2 (Norgaard et al. 2012). Consequently, pesticides seem to leach only from the northern part of the field.

Fig. 1
figure 1

Schematic presentation of the Silstrup field. Sample positions are indicated by the black dots. The horizontal wells (H1 and H2) and the screen sections in each well are indicated by the lines. The arrow indicates the groundwater flow direction

Full size image

2.2 Soil Sampling

The cultivated area of the field in Silstrup covers 1.69 ha. Sixty-five samples were sampled from a 60 × 165 m rectangular field with a distance of 15 m between sampling points (Fig. 1) on 6 December 2011. Soil was sampled from the plough layer at a depth of approximately 8–16 cm. First, the upper 8-cm top soil was removed and then a sample was taken by pounding a sterile 50-ml centrifuge tube (upside down) into the ground until the tube was almost full and then the tube was sealed. In the lab, each sample was homogenized by sieving twice through a sterile 4-mm mesh. The soil was further mixed thoroughly and stored at 2 °C for 1 week. Bulk soil for physical and chemical soil analyses were sampled from the top soil on 10 October 2010 (Norgaard et al. 2012).

2.3 Physical and Chemical Soil Analyses

Soil texture was determined according to Gee and Or (2002) using a combined sieve/hydrometer method. Organic carbon was determined on a LECO analyzer coupled with an infrared CO2 detector (Thermo Fisher Scientific, Inc.). Bulk density was determined from weights of 20 × 20 cm (6283 cm3) intact soil columns after drying at 105 °C for 2 weeks (Norgaard et al. 2012). The soil pH was measured in a soil/water suspension of 1:4 (v v −1), and the soil electrical conductivity (EC) was measured in a soil/water suspension of 1:9 (v v −1). Oxalate-extractable iron, aluminum, and phosphorus were determined at AGROLAB GmbH, Germany, using the procedure described by Schoumans (2000).

The Dexter index (Dexter n) for each soil sample was calculated as the ratio (w w −1) between clay and organic carbon (de Jonge et al. 2009; Dexter et al. 2008). The Dexter index is interesting, because it impacts the ability of soil to maintain and regenerate functional structure, satisfactory tilth conditions, clay dispersion, soil-water repellency, sorption, and other physicochemical interphase processes (de Jonge et al. 2009; Dexter et al. 2008).

2.4 Mineralization Potentials

The mineralization potential of the four pesticides was tested using a modified version of a radiorespirometric microplate method (Johnsen et al. 2009). [Aniline ring-U-14C]diflufenican (>98 % purity) was purchased from Chiron A/S (Trondheim, Norway). [Ring-U-14C]bromoxyniloctanoate (99.99 % radiochemical purity) and [p-methylene-14C]glyphosate (>99 % radiochemical purity) were purchased from IZOTOP, Institute of Isotopes (Budapest, Hungary). [CO-14C]bentazone (>98 % radiochemical purity) was purchased from Internationale Isotope (Munich, Germany). Radioactive pesticide solutions (10 mg mL−1, approximately 870 Bq mL−1) were prepared by dissolving appropriate amounts of radioactive pesticide and the corresponding non-labeled pesticide in sterile water (glyphosate and bentazone) or sterile water containing 100 mg L−1 of the Tween-80 detergent (bromoxyniloctanoate and diflufenican). Tween-80 was applied to keep the hydrophobic herbicides in aqueous solution during handling.

For each of the 65 homogenized soil samples, subsamples of 0.5 g were transferred to microplate wells, one microplate for each of the four pesticides and one subsample per pesticide. The microplates were 96-well polypropylene microplates (Nunc 278752) with a well-volume of 2.0 mL to minimize oxygen depletion. Fifty microliters of 14C-labeled pesticide solution was added to all wells, corresponding to an initial pesticide concentration of 1 mg kg−1 soil. The microplates were sealed with PCR sealing tapes on which 96 14CO2 traps (Ca(OH)2-amended filter paper discs) were placed in a pattern corresponding to the microplate wells. Polyurethane foam sheets (the size of a microplate lid) were placed on top of the sealing tapes, microplate lids were added, and the plates and lids were held tightly together with strong rubber bands. The sealing tapes were changed after approximately 2, 3, 6, 10 16, 23, 37, 51, 65, 80, 106, 120, 134, and 148 days of incubation at 10 °C. The trapped 14CO2 from each well, captured on the Ca(OH)2-impregnated filters, was quantified from a standard series of NaH14CO3 using digital autoradiography and subsequent digital image analysis as described by Hybholt et al. (2011).

2.5 Mineralization Kinetics

A two-parameter exponential model (first-order kinetics, Eq. 1) was used to fit the mineralization curves for glyphosate and bromoxyniloctanoate.

$$ y=a\cdot \left(1-{e}^{-bt}\right) $$
(1)

where y is the accumulated 14CO2 (% of added 14C) released at time t (day), a is the maximum 14C mineralized (% of added 14C), and b is the mineralization rate constant (day−1). Since we were interested in estimating the in-situ mineralization potentials, we fitted only the first 23 days of mineralization, where the mineralization followed first-order kinetics. The initial rate at time zero was then calculated from the first derivative function (Eq. 2).

$$ dy/d{t}_0= ba $$
(2)

A linear regression model (Eq. 3) was used to fit the mineralization curves for diflufenican and bentazone.

$$ y=a+bt $$
(3)

For diflufenican and bentazone, the models are based on the mineralization data from days 23–84 and 16–65, respectively. This was done in order to capture the initial, linear part of the mineralization curves from the first detection of mineralization in each of the two cases. The slope of the linear models was used as an estimate of the initial mineralization rate.

2.6 Two-Dimensional Interpolation and Statistical Analysis

The spatial, field-scale variation in soil texture, organic carbon content and the mineralization rates were mapped using minimum curvature interpolation with regularized spline interpolation in ArcMap 10.1. Minimum curvature interpolation generates the smoothest possible surface while still honoring the source data as closely as possible. The number of points used in the calculation of each interpolated cell was set to 12 and the weight parameter to 0.1. The mineralization rates were correlated to soil physical and chemical parameters using the linear correlation coefficient (R 2), as it shows the fraction of the variation in the mineralization potentials that can be explained by the variation in the physical or chemical soil parameters. Coefficients of variation (CVs) for the pesticide mineralization rates and the soil parameters were calculated as the standard deviation divided by the mean and are given as percentage.

2.7 Most Probable Number of Pesticide Degraders

The most probable numbers (MPNs) of cultivable glyphosate-, bromoxyniloctanoate-, diflufenican-, and bentazone degraders were estimated by a modification of the above microplate radiotracer method. To represent the gradients in clay and organic carbon across the field, selected samples were pooled into groups with high clay and low organic carbon content, low clay and high organic carbon content, and intermediate clay and organic carbon content (five to seven subsamples for each group: group A with 17.6–18.9 % clay and 1.8–1.9 % organic carbon, group B with 14.2–14.3 % clay and 2.0–2.1 % organic carbon, and group C with 16.1–16.2 % clay and 1.9–2.0 % organic carbon).

Four grams of the homogenized sample from each of the three groups was shaken for 10 min with 36 mL of half-strength Bushnell-Haas minimal medium with EDTA (magnesium sulfate, 0.1 g; calcium chloride, 0.01 g; monopotassium dihydrogen phosphate, 0.5 g; diammonium hydrogen phosphate, 0.5 g; potassium nitrate, 0.5 g; ferric chloride, 0.025 g; Na2EDTA·2H2O, 6.4 mg; pH = 6.8). These extracts were then threefold serially diluted in half-strength Bushnell-Haas minimal medium. For each pesticide, four subsamples of 500 μL from each dilution were transferred to a deep-well microplate and the wells were spiked with 50 μL of the 14C-labeled pesticide solutions. The plates were then treated as described above for mineralization in the soil samples. A well was considered mineralization-positive if the accumulated amount of 14CO2 at the end of the experiment (148 days at 10 °C) exceeded 5 % of the initially added 14C-labeled pesticide. The MPNs were calculated according to Hurley and Roscoe (1983) from the distributions of positive and negative microplate wells. The lower detection limit was calculated by assuming only one mineralization-positive well at the lowest dilution (10-fold), and the upper limit was calculated from only one mineralization-negative well at the highest dilution (21,870-fold).

3 Results

3.1 Pesticide Mineralization

The MPNs of the microbial degrader populations were for glyphosate and bromoxyniloctanoate above the detection limit of 6.6 × 104 cells g−1 soil, which indicates a large potential for microbial degradation of these pesticides. This was reflected in the rapid mineralization without any-lag phase of both glyphosate and bromoxyniloctanoate (Fig. 2). After a fast, immediate phase, the mineralization of glyphosate leveled off at 10–20 % and bromoxyniloctanoate at 13–26 %.

Fig. 2
figure 2

Mineralization curves depicting the variation in mineralization within the field. The curves represent for each herbicide the two soil samples with the lowest initial mineralization rate and the two with the highest

Full size image

Diflufenican and bentazone both showed slow linear mineralization with a lag-phase, and both pesticides reached very low mineralization levels (diflufenican 1–5 %, bentazone 3–7 %) within the 148 days of the experiment (Fig. 2). The first mineralization was detected on day 23 for bentazone and on day 37 for diflufenican. We did not detect any microorganisms that could utilize diflufenican or bentazone as a source of carbon and energy (MPN <4 cells g−1 soil), which probably explains the long lag-phases in the mineralization of these two herbicides.

The mineralization of bromoxyniloctanoate showed the best model fits (R 2 = 0.980–0.996, average 0.992, Fig. 3), whereas the glyphosate mineralization was slightly underestimated within the first 3 days and slightly overestimated the following 13 days (R 2 = 0.933–0.987, average 0.968). The model fits for diflufenican mineralization (R 2 = 0.734–0.995, average 0.963) and bentazone mineralization (R 2 = 0.850–1.00, average 0.992) were more variable.

Fig. 3
figure 3

Examples of herbicide mineralizations and the corresponding model fits. The data represent for each herbicide the two soil samples with the lowest initial mineralization rate and the two with the highest initial mineralization rate within the fitted time period

Full size image

3.2 Field-Scale Variation in Pesticide Mineralization Rates

The spatial variability of the initial mineralization rates, derived either from the initial rate at time zero for glyphosate and bromoxyniloctanoate or as the slope of the linear regression models for diflufenican and bentazone, is depicted in Fig. 4. Throughout the field, the initial glyphosate mineralization rates varied from 12.1 to 26.0 μg kg−1 day−1 (average 17.1 μg kg−1 day−1, CV = 16.7 %), with a slight indication of lower mineralization in the southern part of the field. Bromoxyniloctanoate had the largest initial mineralization rates varying from 14.9 to 42.0 μg kg−1 day−1 (average 29.6 μg kg−1 day−1, CV = 16.5 %). Diflufenican and bentazone showed very limited mineralization of only 0.11–0.58 μg kg−1 day−1 (average 0.32 μg kg−1 day−1, CV = 24.7 %) and 0.13–0.64 μg kg−1 day−1 (average 0.47 μg kg−1 day−1, CV = 22.4 %).

Fig. 4
figure 4

Spatial distributions of herbicide mineralization rate, clay-, silt-, sand-, and organic carbon (OC) content, and Dexter n (clay/OC ratio). The dots denote the sampling points (n = 65)

Full size image

3.3 Mineralization Rates and Soil Characteristics

The range, mean, and CV for the measured soil parameters are reported in Table 2. Gradients in clay and organic carbon content run in opposite directions within the field. Thus, highest clay contents and lowest organic carbon contents were found in the northern part of the field and lowest clay and highest organic carbon contents were found in the southern part of the field (Fig. 4). The ratio between clay and organic carbon, Dexter n, therefore increased from south to north (Fig. 4). The mineralization rates for each of the four pesticides generally showed no correlation or very little correlation to the soil parameters (Table 3). The highest correlation was between the glyphosate mineralization and the Dexter n, but this correlation was also weak (R 2 = 0.17).

Table 2 Basic soil parameters. Minimum, maximum, mean, and coefficient of variation (CV) of soil texture, organic carbon (OC), Dexter n, bulk density, oxalate-extractable aluminum (Al), oxalate-extractable iron (Fe), and oxalate-extractable phosphorus (P), pH, and electrical conductivity (EC)
Full size table
Table 3 The linear correlation coefficients (R 2) between the pesticide mineralization rates and the basic soil parameters
Full size table

Linear correlations between the mineralization rates of the four pesticides are reported in Table 4. As in Table 3, the correlation coefficients are weak and the strongest correlation was between the mineralization rates of bromoxyniloctanoate and bentazone (R 2 = 0.16).

Table 4 The linear correlation coefficients (R 2) between the pesticide mineralization rates
Full size table

3.4 Field-Scale Leaching

Water from the two horizontal wells (H1 and H2, Fig. 1) was analyzed for glyphosate and bentazone and their main metabolites aminomethylphosphonic acid (AMPA) and 2-amino-N-isopropylbenzamide (Table 5). Glyphosate and bentazone represent both types of mineralization kinetics found in the mineralization experiment. Glyphosate was applied five times on the field during the period from 1988 to soil sampling in 2011 with two applications within the monitoring period (2001 and 2003, Table 1). The glyphosate and AMPA contents in water from different subsections of H1 and H2 were analyzed from 2001 to 2005. Glyphosate was not detected in any of the samples. The glyphosate degradation product AMPA, however, was detected in 6.3 % of the analyzed samples from H1.2 and 25 % of the analyzed samples from H1.3 (Fig. 5). None of the AMPA concentrations exceeded the drinking water quality criterion of 0.1 μg L−1. Bentazone was applied four times on the field from 1994 to 2011, and two of these applications were within the monitoring program (2003 and 2009, Table 1). Bentazone was analyzed for in the periods from 2003 to 2006 and from 2009 to 2011. The bentazone metabolite was analyzed for only in the period from 2003 to 2006. In total, bentazone was detected in 20 % of the samples from H1, 19.5 % of the H1.2 samples, and 10 % of the H1.3 samples (Fig. 5). One of the detections in H1.2 was above the criterion of 0.1 μg L−1. The bentazone degradation product, 2-amino-N-isopropylbenzamide, was not detected in any of the analyzed samples. The metabolite, however, was analyzed for only in the period from 2003 to 2006, whereas bentazone was analyzed for in the periods from 2003 to 2006 and from 2009 to 2011. The horizontal monitoring well, H2, was suspended from 2009.

Table 5 Number of groundwater samples from H1 and H2 analyzed for glyphosate, aminomethylphosphonic acid (AMPA), bentazone, and 2-amino-N-isopropylbenzamide
Full size table
Fig. 5
figure 5

Percentage of samples from the two horizontal wells, H1 and H2 (Fig. 1), containing detectable levels of the glyphosate degradation product AMPA or bentazone. Water was collected monthly (H1.2 and H2.2, n = 29–63) or half yearly (remaining filter sections, n = 4–10)

Full size image

4 Discussion

In this study, we have investigated the potential mineralization of four herbicides commonly used in agriculture (Miljøstyrelsen 2014). These herbicides represent different physico-chemical properties with very different literature reports on hydrophobicity and sorption. Glyphosate was an easily mineralized, hydrophilic compound (logK OW = −3.2) (Tomlin 2000) with strong sorption to clay loam (K d  = 344–667 L kg−1) (Paradelo et al. 2015). Bromoxyniloctanoate was also easily mineralized and strongly sorbing (K d  = 196 L kg−1, estimated from K oc) (Wauchope et al. 1992), but hydrophobic (logK OW = 5.4) (Tomlin 2000). Diflufenican was difficult to mineralize, hydrophobic (logK OW = 4.9) (Tomlin 2000), and strongly sorbing (K d  = 108–694 L kg−1) (Benoit et al. 2008; Lecomte et al. 2001), and bentazone was also difficult to mineralize, in spite of being hydrophilic (logK OW = −0.46) (Tomlin 2000) with low sorption (K d  = 0.06 L kg−1) (Rodríguez-Cruz et al. 2006).

It is clear from the above that bioavailability, expressed as the soil/water distribution coefficient K d , did not determine the different mineralization patterns between the four herbicides. One reason could be that we added the Tween-80 detergent to the solutions of bromoxyniloctanoate and diflufenican to be able to handle these compounds in aqueous solution. This is, to a large extent, the same effect as seen for surfactants in the proprietary commercial formulations of these herbicides. Tween-80 may have improved the initial availability of bromoxyniloctanoate, so that it could be quickly hydrolyzed to its more bioavailable bromoxynil form (Collins 1973), whereas Tween-80 seems to have had little effect on diflufenican mineralization. Also, bentazone was not mineralized to any great extent in spite of high bioavailability, which suggest a microbiological limitation rather than a physico-chemical limitation. Glyphosate, in contrast, was easily mineralized in spite of a high distribution coefficient and thus low bioavailability, indicating that sorption may be less important when degraders are very numerous in the soil.

We used first-order kinetics to quantify the mineralization of glyphosate and bromoxyniloctanoate for the first 23 days. Such simple concentration-dependent (first-order) kinetics indicates no-growth conditions, which is consistent with the high initial number of microbial degraders. Similar mineralization curves with immediate concentration-dependent mineralization have been reported previously for both glyphosate (Gimsing et al. 2004), bromoxynil (Zablotowicz et al. 2009), and bromoxyniloctanoate (Rosenbrock et al. 2004). The slow-down of mineralization after the initial first-order mineralization was presumably caused by aging processes, where an increasing fraction of the herbicides may become permanently sorbed in micropores (Alexander 1995) or pesticide-derived 14C may be incorporated in microbial biomass that is subsequently stabilized within the soil organic matter (Nowak et al. 2011).

Linear regression was used to quantify the mineralization of diflufenican and bentazone covering the time periods 23–84 and 16–65 days, respectively. The linear mineralization patterns indicate that these pesticides were probably mineralized by slow co-metabolic metabolism without growth of the degrader organisms, which is consistent with the absence of bacteria that could utilize them for growth. The 2–3-week delay in mineralization may imply that the degrader organisms were fungi. Possibly, the fungal mycelia were negatively affected by the repeated sieving and therefore needed time to recover. Comparable slow, linear mineralization of bentazone was previously reported for a sandy clay loam soil (Rodríguez-Cruz et al. 2006), and Bending et al. (2006) also reported slow linear diflufenican degradation.

Repeated spraying with a pesticide may induce an increased potential for mineralization of that pesticide due to microbial adaptation and growth. The most frequently applied herbicide on the field was glyphosate (Table 1), which might have added to the high mineralization potential for this compound, though the last application was 9 years prior to our soil sampling. There seems to have been little effect of the repeated application of bentazone, applied as late as 2009, which is consistent with the absence of degraders that could utilize this compound for growth in the MPN assay.

We expected that the two gradients in clay and organic carbon would highlight the effects of either clay or organic carbon content on the mineralization potential, but this was not the case. The mineralization potentials of bromoxyniloctanoate, diflufenican, and bentazone did not correlate with the gradients in clay and organic carbon across the field or any other of the measured soil parameters. Only the glyphosate mineralization rates tended to increase towards the northern part of the field, correlating slightly with increasing clay and decreasing organic carbon contents (Table 3). The indication of a positive correlation to clay was surprising as glyphosate often sorbs strongly to clay (de Jonge et al. 2001), which would reduce the glyphosate bioavailability. The highest correlation was, however, between the glyphosate mineralization and Dexter n, so that it was the ratio between clay and organic carbon more than the total contents that influenced the glyphosate mineralization.

The pesticide mineralization rates exhibited higher coefficients of variation than the soil parameters which have also been reported by others (Röver and Kaiser 1999; Vinther et al. 2008). For pesticide degraders, it has been proposed that high spatial variability is typically associated with populations of limited abundance, while more abundant populations are more homogeneously distributed (Dechesne et al. 2014). This was supported in our study where mineralization of glyphosate and bromoxyniloctanoate both showed low CVs (16.7 and 16.6 %) and high-degrader MPNs (>6.6 × 104 cells g−1), whereas mineralization of diflufenican and bentazone showed high CVs (24.7 and 22.4 %) and low-degrader MPNs (<4 cells g−1).

Our results indicate that the development of generally valid models for predicting pesticide mineralization across field sites, based on simple soil characteristics and in-vitro mineralization rates, may be unrealistic. Furthermore, if the mineralization of two or more of the herbicides were determined by the same soil parameters, we would have seen correlations between these herbicides, which were not the case (Table 4).

When comparing our results to the literature, it becomes clear that there are few general trends (if any) for studies that seek to correlate mineralization of the four herbicides to field parameters. For instance, for glyphosate, Vinther et al. (2008) found much higher positive correlation between mineralization and organic carbon (R 2 = 0.56) and between mineralization and clay content (R 2 = 0.24). This was believed to be caused by the strong glyphosate adsorption to clay and humus, leading to a close contact between the glyphosate and the glyphosate-degrading organisms in the soil. For bentazone, Ghafoor et al. (2011) reported strong microbial control on degradation as well as pH dependence, whereas glyphosate degradation showed a strong relationship to the Freundlich sorption coefficient and laccase activity. Rodríguez-Cruz et al. (2006) reported a positive correlation between bentazone degradation and organic carbon, whereas we did not find any correlation between bentazone mineralization and organic carbon. A study by Bending et al. (2006) showed little variation in diflufenican degradation in spite of variation in the measured soil parameters (C, N, pH, and dehydrogenase activity), which is similar to the mineralization results of our study.

Phenoxyalcanoic acid herbicides are another popular group of model herbicides. Fredslund et al. (2008), for instance, studied MCPA mineralization in soil from the Ap and Bs horizons of a sandy agricultural field. The MCPA mineralization at their study site did not correlate to any of the measured soil parameters, which included sorption, soil pH, organic carbon content, clay content, and six microbial parameters. Shymko and Farenhorst (2008) also did not detect any correlation between 2,4-D (2,4-dichlorophenoxyacetic acid) mineralization and clay or soil organic carbon in samples from a cultivated, Canadian prairie landscape, but the mineralization was, to some degree, influenced by soil moisture.

It seems difficult to connect pesticide mineralization (or degradation) and specific topsoil parameters, but what about pesticide mineralization and leaching? Though included in the analyses, we did not detect glyphosate in the samples from the horizontal monitoring wells, but the glyphosate degradation product, AMPA, was detected. In contrast, only bentazone was detected, and not the degradation product, 2-amino-N-isopropylbenzamide. All detections of AMPA and bentazone were from the H1 well that collected water from the northern part of the field. Neither AMPA nor bentazone was detected in the samples from H2 which collected water from the southern part of the field. This pattern does not correspond well with the rather random distribution of mineralization potentials of the two herbicides (Fig. 4). The reason is probably that even a high microbial potential for degradation and mineralization in the plough layer can be of little relevance in macroporous soils, where pesticides may be transported rapidly from the surface to deeper layers after intense precipitation (Borggaard and Gimsing 2008; de Jonge et al. 2000; Kjær et al. 2011; McGrath et al. 2010; Norgaard et al. 2014; Vereecken 2005). Such bypass of the active degrader layer by preferential flow through worm holes was recently demonstrated in detail by high-resolution model simulations of MCPA degradation and transport in a comparable agricultural soil (Rosenbom et al. 2014). Our data seems to support their model predictions.