The spatial and temporal distribution/variation of pesticide residues in Viotikos Kifissos basin before and after the application of a low input crop management system. A three-year study

A three-year study of the spatial and temporal variation of pesticide residues in Viotikos Kifissos basin (VKB), in central Greece, before and after the application of Low-Input Crop Management systems (LCM), was carried out between May 2009 and November 2011. A total of 253 water samples from 34 sampling points (boreholes, rivers, wells, vadose zone) were analysed for the presence of more than 200 pesticides and degradation products, using liquid and gas chromatography coupled to tandem mass spectrometry. Samples were prepared by solid-phase extraction (SPE) using Oasis HLB cartridges. Almost 70% of the samples were found positive for the presence of pesticides in amounts greater than the limit of detection. The most frequently detected pesticide was fluometuron, followed by s-metolachlor, metalaxyl-M, imidacloprid and ethalfluralin. The highest concentration was recorded for fluometuron (322 μg L−1), ethalfluralin (123 μg L−1) and s-Metolachlor (36.4 μg L−1). Results showed a reduction of the pesticides detected and quantified in water samples during 2011 compared to those measured in 2009 and 2010.


Introduction
Extended use of pesticides, either for plant protection or for public health purposes, is one of the main reasons affecting the quality of surface and groundwater bodies, and although sequences on surface water can be seen immediately, on groundwater this can take years. Distribution and fate of pesticides in the environment were well described by Leonard et al. in 1976 [1]. This includes (1) leaching towards surface and groundwater, (2) drift during spraying to long distances, (3) run-off along with soil erosion, and (4) evaporation from sprayed surfaces, transportation through atmosphere and return to land via rain or snow. Various parameters such as weather conditions, soil characteristics, agricultural practices, physicochemical properties and fate in the environment of the used pesticides can affect the amount of sprayed pesticides reaching a destination other than their target species. Also, the incorrect disposal of empty containers and spraying residues can cause point source pollution of surface water. temporal variation and distribution of plant-protecting products, before (first year of the study) and after the application of Low Input Management systems (second and third year), in Viotikos Kifissos basin (VKB) in Central Greece. Water analysis served more like a tool to the whole project in order to evaluate the effectiveness of this reduced input system, so the results from the last two years were not only evaluated as absolute numbers, but also compared with the respective ones from the first year. A total of 253 water samples from boreholes, rivers, wells and vadose zone were analysed for more than 200 pesticides and degradation products with a method based on solid-phase extraction (SPE) using Oasis HLB cartridges followed by LC-MS/ MS and GC-MS/MS analysis. Method's scope included almost all pesticides registered for crops in the studied area, besides glyphosate and any other active substance that required a single residue method for its determination.

Description of the studied area
The project was located in VKB, one of the most productive plains of Greece, and an area in which intensive agricultural practises are followed. The abundance of water reserves in the basin and the intensiveness of the agricultural activities make this agro-ecosystem extremely vulnerable to pollution by pesticides. The region also has a wider socio-economic importance. The basin drains into Lake Yliki, one of the main reservoirs for drinking water for Attiki, which is the largest prefecture in Greece, with a population of about 5 million inhabitants.
The project was applied on a pilot scale in an area of 900 ha of agricultural land, within the VKB and adjacent to Viotikos Kifissos River and Lake Yliki ( Figure 1). The area is cultivated mainly with crops such as cotton, maize and plum tomato, and one of the main challenges of crop protection in the studied area was weed control, especially purple nutsedge (cyperus rotundus).

Hydrogeology of the studied area
The dominant hydrological feature in the studied area is the Viotikos Kifissos River that originates from Mountain Parnassus and ends through an artificial tunnel at Yliki Lake.
VKB spans at 2720 km 2 and consists of three sequential interconnected parts, namely Upper route, Median route and Lower route (where Copais plain is located) including the sub-basin of Yliki-Paralimni. General groundwater flow along the parts of VKB heads from NW to SE. Copais plain is recharged by the lateral cross-flows of the Median Route and the hydraulic connections with mountain Parnassus (SW) and mountain Helikonas (S).
Groundwater flows mainly through the extended karstic massif that connects the adjacent routes of VKB. At the plain parts of the basin, shallow alluvial aquifers exist with significantly smaller potential compared to the karst, still important as a buffer zone (hydrologically and in terms of pollution migration).
Prevalence of shallow levels in alluvial aquifer and the hydrogeological characteristics of the karstic aquifer along with the systematic and intensive use of pesticides make groundwater in the VKB susceptible to contamination.
The karstic aquifer is very vulnerable to pollution due to (1) high permeability, (2) increased velocities of groundwater and (3) low potential of physical attenuation processes. In addition, run-off from cultivation areas and livestock contributes significantly to environmental pollution, and the over-exploitation of groundwater resources leads to water deficiency and deterioration of water quality.

Sampling
A total of 253 samples were collected from 34 sampling points. According to water type, sampling points are described by the letters B (boreholes), R (rivers), V (vadose zone) and W (wells). Borehole samples are representative of karstic aquifer and wells samples of alluvial aquifer. Vadose zone leachates are considered as the most representative samples to identify the effects of the agricultural activities since it is the direct receptor of agricultural inputs.
In 2009 samplings took place during May, July and September and for the years 2010 and 2011 during February, May, July and September to coincide with key periods meaning: (1) prior to sowing and any other agricultural practise, (2) post-sowing and after the first applications of plant protection treatments, (3) after the completion of plant protection treatments and (4) after the end of the harvesting period. Samples were collected in amber glass bottles (2.5 L) and stored at 4°C. Within 3-4 days from arrival at the laboratory, samples were filtered to remove particulate matter, and analysed.

Chemicals and reagents
All analytical standards with purity >98% were bought from Dr Ehrenstofer GmbH and Chemservice. All solvents, namely acetonitrile, ethyl acetate, methanol and water, were of HPLC grade. Oasis HLB cartridges (6 mL, 200 mg) were from Waters (Miliford, MA, USA). The pesticides analysed are listed in Tables 1 and 2. The pesticide class, mode of action, negative logarithm to the base 10 of the octanol water partition coefficient (pK ow ) and vapour pressure (Vp) of each pesticide included in the study are summarised in Table 1

Sample extraction
Extraction was performed by offline SPE based on a previously developed and validated method [22]. Around 500 mL of water samples without pH adjustment were loaded to a previously conditioned (5 mL ethyl acetate, 5 mL methanol) and equilibrated (10 mL water) Oasis HLB cartridge (200 mg, 6 mL), which was attached to a 12-position vacuum manifold, at a flow rate of 15 mL min −1 . Air was passed through the cartridges for 30 min to remove residual water. The cartridges were then disconnected and the adsorbed pesticides were eluted with 5 mL of methanol and 5 mL of ethyl acetate. The eluate was evaporated to dryness using a rotary evaporator at 40°C, and the residue was reconstituted in 1 mL of acetonitrile and injected to LC-MS/MS and GC-MS/MS systems after the filtration using a 0.45 μm syringe filter made of polytetrafluoroethylene (PTFE). Validation data (accuracy, precision and LOQ) are summarised in Tables 2 S and 3S. 2.4. Chromatographic analysis 2.4.1. Liquid chromatography coupled to triple quadropole mass spectrometer The liquid chromatographic (LC) system used consisted of an Agilent1200 Series Quaternary system. Chromatographic separation was achieved using an Eclipse XDB C18 column (15 cm, ID 2.1 mm, 5 μm) at a flow rate of 0.31 mL min −1 using a gradient program with a mobile phase consisting of water-5 mmol L −1 ammonium formate-0.1% formic acid-0.02% acetonitrile (solvent A) and methanol-5 mmol L −1 ammonium formate-0.1% formic acid (solvent B). Initial composition of the mobile phase was 20% of solvent B and 80% of solvent A. This composition was held for 2 min, ramped linearly over the course of 12 min to 60% of solvent B   and ramped linearly again over the course of 30 min to 100% of solvent B. This composition was held for additional 5 min, before returning to the initial condition at 35.1 min. The column was re-equilibrated for 10 min at the initial mobile phase composition. The total runtime was 45 min. The injection volume was 5 μL and in order to avoid carry-over, the autosampler was purged with a mixture of methanol/water (50/50 v / v ) before each sample injection. Mass spectrometer used was the Agilent 6410 equipped with an electrospray ionisation (EI) interface operating in the positive mode. Typical source parameters were as follows: capillary voltage (CV) and collision cell energy varied depending on the precursor ion. Source temperature was set at 250°C, drying gas flow rate at 11 L min −1 , nebulising gas pressure at 45 psi and CV at 3000 V. The MRM experiments were conducted with a dwell time of 50 msec. For instrument control, data acquisition and processing, Agilent MassHunter software version B.01.04 (B84) was used. Retention time (t R ), quantification and qualification ion transitions, CV and collision energy (CE) for data acquisition in LC-MS/MS are listed in Table 1. 2.4.2. Gas chromatography coupled to triple quadropole mass spectrometer GC-MS/MS analysis was performed by a Varian 3800 gas chromatograph connected to a triple quadrupole mass spectrometer (Varian model 1200L). Samples were injected with a CP 8400 autosampler, using a 100 μL syringe, into a 1079 programmed temperature large volume injector (PTV-LVI). A Factor Four Capillary Column VF-5ms (30 m × 0.25 mm ID × 0.25 μm film thickness), with a guard column (fused silica untreated capillary column 5 m × 0.53 mm ID cyano-phenyl-methyl deactivated) from Varian Inc., was used for the chromatographic separation of the compounds. The mass spectrometer was operated at the EI mode and in the MRM mode. Transfer line, manifold and ion source temperatures were at 280, 40 and 250°C, respectively. The initial injector temperature (90°C) was held for 0.75 min and then increased to 300°C at the rate of 200°C min −1 , and held for 5 min, and cooled back to 90°C at the rate of 200°C min −1 and held for 56 min. The injector split ratio was initially 90:10. After 0.75 min, splitless mode was set until 3 min. At 3 min, the split ratio was 50:1, and at 6 min, the split ratio was set to 10:90. The column oven temperature program started from 85°C for 2 min, increased to 180°C at a rate of 30°C min −1 , then increased to 230°C at a rate of 1.8°C min −1 , then increased to 280°C at a rate of 30°C min −1 , and held for 30 min. The helium carrier gas flow rate was 1 mL min −1 . Injection volume was set at 25 μL, and because of the equipment's low sensitivity, desirable analytical limits could not be reached otherwise. Retention time (t R ), quantification and qualification ion transitions, and CE for data acquisition in GC-MS/MS are listed in Table 2.

Confirmation criteria
For initial identification of the analytes, the retention time criterion was used. The t R of the analyte was matched based on an external calibration standard at a tolerance of ±0.2 min for both liquid and gas chromatography. The final confirmation was carried out based on the criterion of the ion ratio where the recommended maximum difference between an external calibration standard and a sample is 30% for both GC-MS/MS and LC-MS/MS systems. These identification criteria are in accordance with the European Union guidelines for the determination of pesticides residues [23].

Treatment of data
The results by the three-year study were statistically analysed using one-way analysis of variance (one-way ANOVA), which tests the null hypothesis; thus samples in two or more groups are drawn from populations with the same mean values. All data were analysed using Microsoft Office Excel 2007 for a 95% confidence level. The ANOVA produces an F-statistic, which is the ratio of the variance calculated among the means to the variance within the samples, and is compared to F-criterion, which is the F-value produced when the null hypothesis is valid. So when the F-statistic is greater than the F-criterion, the null hypothesis is not valid and the samples are drawn from different populations.

Results and discussion
A total of 253 samples were analysed, from 34 sampling points, with 19 samples from boreholes, 30 from rivers, 87 from wells and 117 from vadose zone leachates. Seventy percent of the samples were found positive for the presence of one or more pesticides in a concentration greater than the detection limit. Figure 2 shows the distribution of the samples and the number of positive samples per category. The most frequently detected pesticide was herbicide fluometuron with a percentage of 55.9% over the positive samples, followed by s-metolachlor (50.3%), metalaxyl-M (40.1%), imidacloprid (31.6%), ethalfluralin (19.8%), terbuthylazine (18.1%), trifluralin (15.8%) and desethylterbuthylazine (DET) (10.6%). The highest concentration was recorded for fluometuron (322 μg L −1 ), ethalfluralin (123 μg L −1 ) and s-Metolachlor (36.4 μg L −1 ). Besides imidacloprid and metalaxyl, all other pesticides regularly detected were herbicides. Concentrations and high detection frequency can be assigned to the wide use of herbicides, since as already mentioned, one of the main challenges of crop protection in the studied area was weed control. The presence of herbicide residues Physicochemical properties and GUS index values of herbicides like fluometuron and therbuthylazine, indicate that they are potential leachers (GUS value >2.8) and that is possible to be detected in groundwater.
DET, the degradation product of herbicide terbuthylazine, was detected in 19 samples, 18 of which were from vadose zone and one sample from wells. DET was detected along with terbuthylazine, besides four cases, where terbuthylazine was not detected above LOD, indicating that it hadn't been used in the current applications.
Our results were in accordance with the results from other studies. Terbuthylazine and its metabolite DET were detected in 96% of groundwater samples analysed from a vineyard region in Spain, even in concentrations 10 times higher than the EU limit for groundwater. Other herbicides frequently detected (in more than 50% of samples) were fluometuron and metolachlor. Fungicide metalaxyl was detected in more than 50% of the samples, but insecticides included in the study were detected in much smaller numbers of samples [24]. Atrazine, fluroxypyr, metolachlor and terbuthylazine were detected frequently in samples from wells used for irrigation purposes in central Italy. The detected average concentration was below 0.1 μg L −1 , but the highest concentration recorded was that of metolachlor at 12.5 μg L −1 [25]. Pesticides widely used in Mediterranean agriculture such as diazinon, imidacloprid, metolachlor and terbuthylazine were found in surface water samples taken from Jucar, Ebro, Llobregat and Guadalquivir rivers [26].
Elevated concentrations of terbuthylazine and high detection frequency are a result of atrazine withdrawal from the European market and the use of terbuthylazine as a substitute [27,28].
Among the fungicides studied in our project, metalaxyl-M had the highest detection frequency (40.1% over positive samples), whereas the detection frequency for others (azoxystrobin, benalaxyl, myclobutanil) was below 3%.
With respect to fungicides, metalaxyl was determined in 27% of the surface water samples collected from 10 sites in eight states of the US, with a maximum concentration of 1.15 μg L −1 , whereas azoxystrobin was detected in 50% of the samples [29]. In a different study, metalaxyl was detected in more than 50% of the groundwater samples [24]. Ethalfluralin and trifluralin, although considered as non-leachers according to the GUS index (0.45 and 0.13, respectively), were both among pesticides identified in groundwater and surface water. This was in accordance with other similar studies, and their occurrence can be seen as a result of their constant use in the areas where the sampling points were situated [6,30].
Insecticides azinphos ethyl, chlorpyrifos and pirimiphos methyl were only detected in vadose zone leachates in 2010 (samplings in July and September), when cultivations were infected by Heliothism Armigera, a lepidopter, which in early May attacks cultivations like corn, cotton and tomato. Insecticide imidacloprid, also used against Heliothis Armigera, Lyonetia specculella and Aphis spiraecola, had a maximum annual average concentration in 2010, whereas for 2009 and 2011, the results were mainly below LOD.
Other pesticides detected but not on a regular basis were azinphos ethyl, azoxystrobin, benalaxyl, chlorpyrifos, dinitramine, fluazifop, metribuzin, myclobutanil, pendimethalin, pirimiphos methyl, prometryn, quizalofop, simazine and sulcotrione. Results for these compounds were not included in statistical analysis due to the low detection frequency.
A synopsis of the results is given in Table 3, while in Table 4 all sampling points are listed and information is given for each sampling period.

Seasonal distribution/variation
Maximum concentrations were recorded in spring and summer, which was rather expected since the applications for the crops mentioned already take place late in spring (around May) and early in summer. Lower concentrations in autumn and winter can also be attributed to increased precipitation, which results in dilution and enhanced degradation of applied pesticides. Exception observed was for ethalfluralin (Figure 3), which showed a maximum concentration in spring, but the concentration recorded in winter was greater than the concentration recorded in summer. This may be attributed to a greater run-off observed in winter because of increased precipitation, which overcomes the results of dilution. Even so, the average of spring/summer concentration was again greater than autumn/winter.
DAVLIA P N N P P -N N P P P P V07BOI001 AG. VLASIOS -P P P -P P P P P P P V07BOI002 AG. VLASIOS -N P P P P -P P P - AG. VLASIOS -P P P P -P P P P P P V07BOI005 CHERONIA CHERONIA -N P P P P P P P P --V07BOI007 CHERONIA-AKONTIO -N P P --P P -P P P V07BOI008 CHERONIA -N P P --P P P P P P V07BOI009 CHERONIA -N P P P P P P P P P P V07BOI010 THOURIO -N P P P P P P - THOURIO P N N P P -P P P P P N W07BOI008 THOURIO P P N P P -P P P P P P W07BOI010 THOURIO P P N P P -P P P P P W07BOI016 CHERONIA N N P P P -P P P N P P W07BOI023 Notes: N: no pesticide residues were detected above LOD.
-: no sample was taken. P: pesticide residues were detected above LOD.
One-way ANOVA was used to compare seasonal variation. Results showed no significant difference since for every case checked, F criterion was greater than F statistic at a 95% confidence level.

Annual distribution/variation
Mean annual concentrations for every possible combination pesticide/sampling point were calculated for all three years. Results (Table 5) showed a reduction in absolute numbers for 2011 (third year of the project) compared to 2009 and 2010, for about 70% of the studied cases, meaning that the low input system had a result. One-way ANOVA showed no significant difference though, indicating that the two-year period of program implementation is a short time to change the environmental quality status of an area where intensive agricultural practises  were followed for many years. In only 13% of the cases studied, increased annual average concentrations were observed in 2011 compared to the previous years. The case of sampling point R1 showed that the river had a good quality status, and that the findings were a kind of point source contamination. Imidacloprid had a constant profile, meaning that 2010 had the maximum annual average concentration for all sampling points, compared to 2009 and 2010, which was due to infection of crops by Heliothis Armigera.

Spatial distribution/variation
Samples were collected from eight different towns in the area of VKB and the results about the percentage of positive samples per area are briefly described in Figure 4. Mayroneri Orchomenos and Akontio are three areas that appear to have a lesser amount of samples positive for pesticide residues, and that is because borehole water samples were mainly taken from these areas.

Borehole samples
Samples from boreholes were very limited (18 out of 253), and were mainly taken in the first year of the study, just to monitor the condition of the karstic aquifer. Samples were limited because, due to the program's duration, the effects of the changes in agricultural practices wouldn't be noticeable on karstic aquifer. Results indicate a good quality condition, since just only one sample out of 18 was positive for the presence of pesticides (ethalfluralin, 0.55 μg L −1 ).

River samples
Samples from Viotikos Kifissos river were also limited (30 out of 253) and the results can only indicate and not asses the quality of the river. Limited number of samples was mainly based on the fact that surface water is a dynamic system that changes continuously so the results can mainly indicate the condition of the system at the moment the sample is taken. Out of the 30 samples, 17 were found positive for pesticide residues, at a concentration greater than the detection limit, and six of these had multiple findings. Pesticides detected were ethalfluralin (four samples), fluometuron (10 samples), pendimethalin (one sample), prometryn (three samples), s-metolachlor (four samples) and trifluralin (four samples). Concentrations and frequency of detection showed that these were a result of bad practices and a random incident.

Samples from wells
Samples from wells indicate the quality status of alluvial aquifer, which extends beyond the limits of the pilot and the wider area, and hence the quality status of groundwater is not exclusively related to the activities within the study area. Nevertheless, results showed that the alluvial aquifer was affected by the long-term use of pesticides in the area since 64% of the samples were found positive for pesticide residues. DET, dinitramine, ethalfluralin, fluometuron, imidacloprid, metalaxyl-M, prometryn, s-metolachlor, terbuthylazine and trifluralin were the pesticides detected and quantified. One-way ANOVA using mean concentrations of main pesticides detected (ethalfluralin, fluometuron, imidacloprid, metalxyl, prometryn, metolachlor, terbuthylazine and trifluralin) for each well sampling point showed that in some cases there is no significant variation between the concentration of main pesticides detected in different wells, while for others there is. Difference was observed for ethalfluralin, trifluralin and s-metolachlor, which is a result of frequency detection, and thus use from farmers, and the big variance between recorded concentrations.

Vadose zone samples
Owing to the project's duration, potential noticeable changes in the concentrations of the environmental critical parameters are recorded mainly in the vadose zone leachates, which is the direct receptor of agricultural inputs. For that reason majority of the samples were collected from the vadose zone. Around 88% of samples were found positive for pesticides residues. Twenty-one different pesticides (azinphos ethyl, azoxyxtsrobin, benalaxyl, chlorpyrifos, DET, ethalfluralin, fluazifop, fluometuron, imidacloprid, metalaxyl-M, metribuzin, myclobutanil, pendimethalin, pirimiphos methyl, prometryn, quizalofop, simazine, s-metolachlor, sulcotrione, terbuthylazine and trifluralin were detected in vadose zone samples, indicating the use of them by farmers during the project. Two of them (fluazifop and metribuzin) were only detected in the first year's samples, and thus before the implementation of LCM.
One-way ANOVA using mean concentrations of main pesticides detected (ethlafluralin, fluometuron, imidacloprid, metalxyl, prometryn, metolachlor, terbuthylazine and trifluralin) for each vadose zone sampling point showed no significant difference (with an exception being terbuthylazine), which implies the wide use of these pesticides in the area.

3.4.
Compliance with environmental quality standards (EQS) 3.4.1. Groundwater Out of 35 samples with a single pesticide detected, 15 samples exceeded the limit of 0.1 μg L −1 for groundwater for insecticides ethalfluralin, prometryn, s-metolachlor and trifluralin. Respectively, out of 22 samples with multiple findings (all wells), 19 exceeded the limit of 0.5 μg L −1 , with the sum varying from 0.55 to 123.1 μg L −1 .

Surface water
Regarding surface water samples, the only pesticide included in priority substances according to the Directive 2013/39/EU is trifluralin with a limit set at 0.03 μg L −1 .Trifluralin was only detected three times at concentrations of 0.26, 0.006 (concentration between LOD and LOQ) and 0.046 μg L −1 in three samples collected from two different sampling points.

Conclusions
A total of 253 water samples from boreholes, rivers, wells and vadose zone leachates from VKB in Central Greece were analysed for the presence of more than 200 pesticides and degradation products.
Twenty-one different pesticides with different modes of action (herbicides, fungicides, insecticides) belonging to different chemical groups were detected in a concentration above the limit of detection. The most frequently detected pesticide and with the highest concentration recorded was fluometuron, followed by s-metolachlor and metalaxyl.
Almost 64% of the samples collected from wells were found positive for the presence of pesticides, indicating that alluvial aquifer was affected by the long-term use of pesticides in the basin, since for years now intense agricultural practises were followed.
Karstic aquifer, the status of which was reflected by borehole samples, was in good quality since only 5% of the samples were found positive. Elevated concentrations and number of positive samples for pesticide residues were recorded in vadose zone leachates, which is the direct receptor of agricultural inputs.
The implementation of a Low Input Crop Management system in the last two years of the project had a beneficial impact since a reduction of pesticides detected and quantified was observed in absolute numbers, although one-way ANOVA indicated no statistically significant change.

Disclosure statement
No potential conflict of interest was reported by the authors.

Supplemental data
Supplemental data for this article can be accessed at http://dx.doi.org/10.1080/03067319.2015.1090564.