Pesticide leaching at TYLSTRUP
Chapter 2, PLAP - Monitoring results May 1999 - June 2002
2.1 Materials and methods
2.1.1 Site description and monitoring design
Tylstrup is located in northern Jutland (Figure 1). The test field covers a cultivated area of 1.1 ha (70 x 166 m) and is practically flat, with a windbreak bordering the eastern and western sides. Based on two soil profiles dug in the buffer zone around the test field the soil was classified as a Humic Psammentic Dystrudept (Soil Survey Staff, 1999). The topsoil is characterized as loamy sand with 6% clay and 2.0% total organic carbon (Table 1). The aquifer material consists of about 20 metres of marine sand sediment deposited in the Yoldia Sea. The southern part is rather homogeneous, consisting entirely of fine-grained sand, whereas the northern part is more heterogeneous due to the intrusion of several silt and clay lenses (Lindhardt et al., 2001). During the monitoring period the groundwater table was 3- 4.5 m b.g.s. The overall direction of groundwater flow was towards the west (Figure 2). A brief description of the sampling procedure is provided in Appendix 2. The monitoring design and test site are described in detail in Lindhardt et al. (2001) and the analysis methods in Kjær et al. (2002).
2.1.2 Agricultural management
The 1999 crop was potato for starch production. The cultivar used was Dianella, which is a commonly used variety. During the growing season the field was sprayed with the herbicides linuron and metribuzin and with the fungicide mancozeb. Potassium bromide tracer was applied on 27 May. The potatoes were harvested on 20 October. The yield of tubers was 475 hkg/ha (24% dry matter), which is somewhat less than the average for the location. During the autumn of 1999 the field was disc harrowed several times in order to reduce problems of waste potatoes in the subsequent crop.
The 2000 crop was spring barley (cv. Bartok), which emerged on 10 April. On 13 May, when the crop had 3 tillers, it was sprayed with the herbicide triasulfuron. Stem elongation and heading began in mid May and June, respectively. Combined fungicide and insecticide spraying was carried out on 19 June, in the middle of heading, using propiconazole, fenpropimorph and pirimicarb. The crop was harvested on 21 August yielding 73.3 hkg/ha of grain (85% dry matter) - somewhat above the average for that year and location.
The 2001 crop was winter rye (cv. Dominator), which emerged on 7 October. On 2 November, when the crop had 2 leaves, it was sprayed with the herbicides tribenuron methyl and pendimethalin. Spraying of fungus was done twice on 14 May and 13 June using propiconazole and fenpropimorph. At harvest on 28 August the grain yield was 63.6 hkg/ha. The winter rye was harvested later than usual due to rainy conditions in August.
The 2002 crop was winter rape (cv. Artus). Due to the rainy conditions in August, sowing had to be postponed until the 3 September, more than 2 weeks later than normal. Clomazone was sprayed to combat weeds 2 days after sowing, i.e. before the crop emerged. On 16 October, when the crop had 4 unfolded leaves, weeds were sprayed with clopyralid. Due to the unusually warm weather in October, the rape was well developed at the onset of winter. At the end of 2001, temperatures dropped below zero, and on 1 January the field was covered with approx. 15 cm of snow. At the beginning of January, temperatures rose above zero and on 8 January there was 1-2 cm of standing meltwater at both ends of the field. One week later, all of the water had infiltrated. The rape was fertilized once on the 22 March using commercial fertilizer. The crop was irrigated three times between 24 April and 31 May. The yield of rapeseed was just 26 hkg/ha at 91% dry mater, the low yield being attributable to the late sowing time. Management practice at the site is detailed in Appendix 3, Table A3.1.
Figure 2. Overview of the Tylstrup test site. The innermost white area indicates the cultivated land, while the grey area indicates the surrounding buffer zone. The positions of the various installations are indicated, as is the direction of groundwater flow (by an arrow).
Figure 3. NE-SW cross section based on wells at the Tylstrup site (Lindhardt et al. , 2001). The location of the wells is indicated in Figure 2.
2.1.3 Model set-up and calibration
The MACRO model was applied to the Tylstrup site covering the soil profile to a depth of 5 m b.g.s., always including the groundwater table. The model was used to simulate the water and bromide transport in the unsaturated zone during the full monitoring period May 1999- June 2002.
The model was calibrated to the observed groundwater table measured in the piezometers located in the buffer zone, as well as to measured time series of soil water content at three different depths (25, 60 and 110 cm b.g.s.) from the two profiles S1 and S2 (see Figure 13) and to the bromide concentration measured in the suction cups located 1 and 2 m b.g.s. The calibration procedure involved adjustment of the empirical BGRAD parameter regulating the boundary flow and selected crop and hydraulic parameters. The parameter ASCALE, which is related to the solute exchange between matrix and macropores, was also calibrated, but this had very little effect on the results. Dispersive parameters were not adjusted. For a detailed description of data acquisition, model set-up and calibration procedures, see Kjær et al. (2002).
In recent years there has been some discussion in Denmark regarding field and catchment scale water balance calculations following several investigations that revealed problems with the water balance. Plauborg et al. (2002) examined the problem focussing on precipitation correction factors and calculation of potential evapotranspiration. Based on their recommendations, we made a thorough analysis of the precipitation correction factors used in the PLAP. It was concluded that the precipitation correction factors suggested by Allerup and Madsen (1979) were the most representative for the rain gauges at the PLAP sites. The precipitation corrections applied in Kjær et al. (2002) were therefore replaced by the monthly corrections of Allerup and Madsen (1979). These changes slightly increased the precipitation input, resulting in minor changes in the water balance as compared to the 2001 PLAP report.
2.2 Results and discussion
2.2.1 Soil water dynamics and water balances
In general the model simulations were consistent with the observed data, thus indicating a good model description of the overall soil water dynamics in the unsaturated zone. The calibrated model provides a good simulation of the measured fluctuations in the groundwater table. The dynamics is captured, whereas the amplitude of the fluctuations is less well described. The overall trends in soil water content could be modelled successfully, with the model capturing soil water dynamics at all depths (Figure 4E).
An annual water balance is determined for each monitoring year (July to June; Table 2). Because monitoring at the Tylstrup site was initiated in May 1999, the two months preceding the monitoring year are included as a separate period.
Figure 4. Soil water dynamics at Tylstrup: Locally measured precipitation and simulated percolation 1 m b.g.s. (A), simulated and measured groundwater level (B), and simulated and measured soil water saturation (SW sat.) at three different soil depths (C, D and E). The measured data in B derive from piezometers located in the buffer zone. The measured data in C, D and E derive from TDR probes installed at S1 and S2 (see Figure 2).
Table 2. Annual water balance for Tylstrup (mm/y). Precipitation is corrected to the soil surface according to the method of Allerup and Madsen (1979).
All 3 monitoring periods (July-June) were wet at Tylstrup, with precipitation input exceeding the yearly normal by 15-39%. Except for the first exceptionally wet year, the precipitation input was counterbalanced by the actual evapotranspiration during the summer months (Figure 4A). Generally, percolation 1 m b.g.s is continuous from September to May.
2.2.2 Bromide leaching
In the unsaturated zone the breakthrough of bromide at 1 m b.g.s. started in August 1999, three months after application. The bromide concentration peaked in September, and the leaching continued throughout the whole winter period until March 2000 (Figure 5). As expected, the breakthrough of bromide at 2 m b.g.s. was delayed by a few months, and the concentration profile at this depth was somewhat wider due to hydrodynamic dispersion.
The model is generally able to satisfactorily simulate the bromide transport, and hence also the percolation. In terms of timing and concentration level of the breakthrough curves the bromide transport was well captured by the model. The simulated breakthrough at 1 m b.g.s. is initiated too soon, however. Attempts have been made to delay this initial breakthrough, but without success. The accelerated breakthrough is probably due to the wet months of May and June 1999, and resultant overestimation of percolation deeper than 1 m b.g.s. and therefore of transport of the bromide to this depth. At 1 m b.g.s. bromide is detected 2-3 months longer than simulated by the model. Improved modelling of the latter would necessitate thorough calibration of the dispersivity and mixing layer. The pulse at 2 m b.g.s. is very well described.
A mass balance for the applied bromide tracer based on daily, simulated values of water flux and bromide concentration revealed that by the end of December 1999, 99% of the applied bromide had leached from the root zone (1 m b.g.s.). Considering the measured bromide concentrations (Figure 5), the tail of the main pulse continued throughout January and February 2000, and small amounts of bromide continued to leach as late as autumn 2000. These findings indicate that a minor part of the bromide had diffused into less accessible pore water, which cannot be described by the MACRO model. The overall conclusion, though, is that the applied bromide leached out of the root zone (1 m b.g.s.) within a year of application.
In the saturated zone, marked breakthrough of bromide was detected in all downstream monitoring wells, although the temporal evolution varied markedly within the various monitoring wells (Figure 6 ). A rapid breakthrough of bromide occurred in monitoring well M4, where an elevated bromide concentration was detected just 6 months after application.
Figure 5. Simulated (solid line) and measured (crosses) bromide concentration in the unsaturated zone at Tylstrup. The measured data derive from suction cups installed 1 m b.g.s. and 2 m b.g.s. at locations S1 and S2 indicated in Figure 2. The green vertical line indicates the date of bromide application.
The breakthrough in the other monitoring wells occurred much later, thus indicating a much slower bromide transport, especially in the northern part of the field site. The bromide transport to M2 was thus delayed about a year as compared to M4 (Figure 6). The difference between the various monitoring wells demonstrates the marked heterogeneity within the test field. Silt lenses identified in the northern part of the area probably cause heterogeneous water flow (Kjær et al., 2002; Lindhardt et al ., 2001). Slightly elevated bromide concentrations were detected in monitoring well M1. As M1 is located about 3 m upstream of the treated area, the tracer bromide was not expected to reach it. However, the silt lenses might have deflected the vertical transport through the unsaturated zone, enabling bromide to reach this upstream monitoring well.
During the 3-year monitoring period the majority of the applied bromide seems to have passed the downstream monitoring wells. Other than at M2, decreasing bromide concentrations were observed at the end of the monitoring period in all the downstream monitoring wells. The overall transport time from field application until the majority of the bromide has passed the monitoring wells ranges from 1.5 to 3 years in M4, M3 and M5, respectively.
Bromide concentration measured 6-7, 7-8 and 8-9 m b.g.s. derives from three additional screens installed near M4 and M5 in August 2001 (Figure 6). Finally, it should be noted that based on the bromide concentration detected during the period 1 May 1999-1 November 1999, the background concentration of bromide at Tylstrup was 0.23 ±0.06 mg/l.
Figure 6. Bromide concentration in the groundwater at Tylstrup. The data derive from monitoring wells M1- M7. Screen depth is indicated in m b.g.s. The green vertical line indicates the date of bromide application.
2.2.3 Pesticide leaching
At Tylstrup, the monitoring encompassed 11 different pesticides and several degradation products applied throughout four growing seasons as indicated in Table 3 and Figure 7. It should be noted that precipitation in Table 3 is corrected to the soil surface according to Allerup and Madsen (1979), whereas percolation (1 m b.g.s.) refers to accumulated percolation as simulated with the MACRO model. It should also be noted that as mancozeb (applied here as Dithane DG) and tribenuron methyl (applied here as Express) degrade rapidly, the leaching risk is more associated with their respective degradation products, ETU and triazinamin-methyl. For the same reasons it is the degradation products and not the mother compounds that are monitored in the PLAP (Table 3).
Table 3. Pesticides analysed at Tylstrup with the product used shown in parentheses. Degradation products are in italics. Precipitation and percolation are accumulated from the date of first application (App. date) until 1 July 2002. 1 st month percolation refers to accumulated percolation within the first month after application. Cmean refers to average leachate concentration at 1 m b.g.s. The number of pesticide-positive samples is indicated in parentheses.
Figure 7. Pesticide application, precipitation and irrigation (primary axis) together with simulated percolation 1 m b.g.s. (secondary axis) at Tylstrup. Pesticides applied later than April 2002 are not included.
The leaching risk of pendimethalin, triazinamin-methyl (degradation product of tribenuron methyl), clomazone and clopyralid, which were applied in 2001/2002, will not be evaluated until the 2003 monitoring results become available, i.e. when two years of monitoring data have been collated. It should be noted, though, that apart from two samples containing 0.016 and 0.72 µg/l of clomazone, none of these three pesticides or their degradation products listed in Table 3 have yet been detected in any of the water samples analysed.
None of the pesticides triasulfuron, pirimicarb, propiconazole, fenpropimorph or their degradation products listed in Table 3 were detected in any of the water samples. All the pesticides were applied in summer 2000, during which time precipitation input was close to normal and was almost counterbalanced by actual evapotranspiration, resulting in little percolation during the first month after application (Figure 7 and Table 3). Monitoring of propiconazole and fenpropimorph has not yet been completed, however, but will continue throughout the next monitoring period, thereby providing two years of monitoring data for evaluation of both applications.
With ETU (degradation product of mancozeb) and linuron, the leaching risk was found to be negligible at the Tylstrup site. Linuron was not detected in any of the water samples, whereas ETU was detected in just six samples taken from the unsaturated zone (Kjær et al., 2001) and two samples from the saturated zone in concentrations of 0.02 µg/l. For further information see Kjær et al. (2002).
Metribuzin was only detected in concentrations of 0.01-0.02 µg/l in three water samples. However, two degradation products of metribuzin leached from the root zone (1 m b.g.s.) in average concentrations exceeding 0.1 µg/l. Leaching was most pronounced with metribuzin-desamino-diketo, reaching an annual average concentration of 0.9 µg/l in suction cup S1. Metribuzin-diketo, the other degradation product of metribuzin, also leached, in this case reaching an average concentration of 0.3 µg/l. Both compounds leached throughout the entire monitoring period, and average concentrations exceeding 0.1 µg/l were detected as long as three years after application. Over the 3-year period as much as 1113% and 4-6% of the applied dosage leached as metribuzin-desamino-diketo and metribuzin-diketo, respectively (Table 4 and Figure 8).
Table 4. Estimated average concentration (µg/l) of metribuzin-desamino-diketo and metribuzin-diketo 1 m b.g.s. at Tylstrup. Leached mass refers to the total mass (% of applied metribuzin) leached during the entire monitoring period (1.7.99-30.6.02). The primary data are given in Appendix 5.
Figure 8. Bromide and pesticide concentrations in the unsaturated zone at Tylstrup. The measured data derive from suction cups installed 1 m b.g.s. and 2 m b.g.s. at locations S1 and S2 indicated in Figure 2. The red vertical line indicates the date of bromide application.
The average concentration of pesticides (Table 4) was estimated using the measured pesticide concentration and estimated percolation on a monthly basis. Measured pesticide concentrations were thus assumed to be representative for each sample period, and accumulated percolation rates calculated using the MACRO model were assumed to be representative for both suction cups S1 and S2. It should also be noted that for 1999/2000 the average concentration is given as a range due to the high level of uncertainty that characterized the first analyses in 1999. The primary data and further information concerning the calculation methods are given in Appendix 5.
In the saturated zone, elevated concentrations of metribuzin-diketo were detected in M1, M3 and M4, while the pesticide concentration in the other wells (M5, M6) could not be distinguished from the background level (Figure 9- Figure 11). At Tylstrup, pesticide application prior to the monitoring period has thus resulted in marked groundwater contamination with the degradation products of metribuzin.
Evidence of previous contamination is provided by the initial screening analysis. The degradation products were present in the groundwater before the monitoring started in May 1999 (Kjær et al., 2002). In M3 and M5, both degradation products were detected long before the bromide had reached the monitoring wells. Bromide was not detected in M6, and the marked contamination was thus due to prior application of the pesticide on the neighbouring field located just south of the test site or on the fields located upstream of M6 (Kjær et al., 2002). The two degradation products of metribuzin were also detected in M1. In view of the slightly elevated bromide concentration detected in M1 (Section 2.2.2), part of the water infiltrating the test site might reach M1. Moreover, metribuzin was applied to the neighbouring field located upstream of the test site in 1999 concomitantly with application on the test site. The elevated concentration of degradation products detected in M1 may thus derive from the test site or the upstream neighbouring field. Previous application of pesticides at the test site and neighbouring upstream fields is detailed in Kjær et al. (2002).
The high background concentration found in all monitoring wells makes it difficult to determine whether the elevated concentrations observed in downstream monitoring wells are due to the metribuzin applied during the PLAP or to metribuzin applied on the test site or on the "upstream" fields prior to the PLAP. Consequently it is not possible to fully verify the impact of the metribuzin applied during the PLAP on the quality of the groundwater. It should be noted, though, that the average concentration of metribuzin-diketo in the Tylstrup groundwater was 0.15 µg/l, and that the average concentration exceeded the maximum allowable concentration (0.1 µg/l) at 81% of the screens analysed. High concentrations were also detected in the deep screens installed in August 2001 near M4 and M5 (Section 2.2.1). In fact, the average concentration in the deepest screen located 8-9 m b.g.s. was 0.32 µg/l in M4 and 0.22 µg/l in M5 (Figure 10-Figure 11).
Metribuzin-desamino-diketo was also detected in 57% of the analysed groundwater samples. Apart from one sample reaching 0.14 µg/l concentrations never exceeded 0.1 µg/l.
Figure 9. Bromide and pesticide concentrations in the groundwater at Tylstrup. The data derive from monitoring wells M1 (A,B,C) and M3 (D,E,F). Screen depth is indicated in m b.g.s. The green vertical line indicates the date of application.
Figure 10. Bromide and pesticide concentrations in the groundwater at Tylstrup. The data derive from monitoring well M4. Screen depth is indicated in m b.g.s. The green vertical line indicates the date of application.
Figure 11. Bromide and pesticide concentrations in the groundwater at Tylstrup. The data derive from monitoring well M5. Screen depth is indicated in m b.g.s. The green vertical line indicates the date of application.
Figure 12. Bromide and pesticide concentrations in the groundwater at Tylstrup. The data derive from monitoring well M6. Screen depth is indicated in m b.g.s. The green vertical line indicates the date of application.
The risk of pesticide leaching at Tylstrup can be summarized as follows:
- With triazinamin-methyl (degradation product of tribenuron methyl), fenpropimorph, propiconazole, pendimethalin, clomazone and clopyralid the leaching risk will not be evaluated until the 2003 monitoring results become available, i.e. when a total of two years of monitoring data have been collated. It should be noted, though, that none of these pesticides or the degradation products fenpropimorphic acid, propanamideclomazone, pirimicarb-desmethyl or pirimicarb-desmethyl-formamido have yet been detected in any of the water samples analysed.
- With triasulfuron, pirimicarb, ETU (degradation product of mancozeb) and linuron, the leaching risk was found to be negligible.
- Two degradation products of metribuzin (metribuzin-desamino-diketo and metribuzin-diketo) were found to leach from the root zone (1 m b.g.s.) in average concentrations exceeding 0.1 µg/l. The estimated leachate concentrations of metribuzin-desaminodiketo and metribuzin-diketo reached 0.9 µg/l and 0.3 µg/l, respectively.
- The monitoring results indicate marked groundwater contamination with the degradation products of metribuzin. The average concentration of metribuzin-diketo was 0.15 µg/l, and in 81% of the screens analysed the average concentration exceeded the maximum allowable concentration (0.1 µg/l). Metribuzin-desamino-diketo was also detected in 57% of the analysed groundwater samples. Apart from one sample, the concentration never exceeded 0.1 µg/l. Whether or not the observed groundwater contamination is due to the metribuzin applied during the PLAP or prior to the monitoring period cannot be determined.