Sunteți pe pagina 1din 11

Arch Environ Contam Toxicol (2013) 64:208–218

DOI 10.1007/s00244-012-9833-9

Influences of Sampling Methodologies on Pesticide-Residue


Detection in Stream Water
Zisheng Xing • Lien Chow • Herb Rees •
Fanrui Meng • Sheng Li • Bill Ernst •
Glenn Benoy • Tianshan Zha • L. Mark Hewitt

Received: 7 July 2012 / Accepted: 22 October 2012 / Published online: 15 November 2012
Ó Her Majesty the Queen in Right of Canada 2012

Abstract Traditional grab sampling (GS) used widely in residues was more accurately captured by hydrograph-
the study of water quality has been found lacking in spatial based sampling, and therefore its use in monitoring pro-
and temporal resolution for pesticide residue monitoring in grams is recommended.
stream water. The objectives of this article are to present a
hydrograph-based sampling approach and compare it with
traditional GS according to sensitivity at temporal and Pesticide contamination in surface and subsurface waters
spatial scales and maximum concentrations of pesticide draining from agricultural land has become a prominent
residues detected in-stream. Data collected from streams research focus in the past few decades (Ernst et al. 2009;
receiving water from three nested watersheds located in Kolpin et al. 2006; Lakshminarayana et al. 1992; Milburn
northwestern New Brunswick, Canada, were used in this et al. 1990; Peruzzo et al. 2008; Smith et al. 1990). It is
study. The results showed that the hydrograph-based well-known that accurately detecting contamination levels
sampling method detected 20 to 30 % more pesticide cases is highly influenced by rainfall-runoff characteristic, rate
than GS for rainfall events causing runoff. Grab sampling and timing of pesticide application, pesticide properties,
significantly underestimated average concentrations of soil characteristics, and crop types (Kanazawa 1989; Willis
pesticide residues by 50 % and maximum concentrations and McDowell 1982; Schulz et al. 1998). However, the
by 1 to 3 orders of magnitude. Using a modified sampler impacts of the sampling approach on the assessments of
design, the spatial and temporal variability of pesticide contamination are often overlooked (Madrid and Zayas
2007).
Because of the high cost and difficulties associated with
Z. Xing  L. Chow  H. Rees  S. Li  G. Benoy analyzing water samples for pesticide residues, collecting
Potato Research Centre, Agriculture and Agri-Food Canada, the minimum number of representative samples is eco-
Fredericton, NB, Canada
nomically and scientifically important to both researchers
Z. Xing  F. Meng and risk assessors. Although regular or irregular grab
Faculty of Forestry and Environmental Management, sampling (GS) is typically used to obtain contamination
University of New Brunswick, Fredericton, NB, Canada
levels (Day 1990; Harris et al. 2000; Kanazawa 1989;
B. Ernst  L. M. Hewitt (&) Laabs et al. 2002; Lakshminarayana et al. 1992; Williams
Aquatic Contamination Research Division, National Water et al. 1995), Ernst (2007, 2009) showed that this sampling
Research Institute, Environment Canada, approach may not correctly capture the concentration and
Burlington, ON, Canada
delivery patterns of pesticide residues in recipient stream
e-mail: mark.hewitt@ec.gc.ca
waters. In fact, sampling was often considered a source of
G. Benoy uncertainty, and traditional sampling frequently had several
Environment Canada, Fredericton, NB, Canada limitations (Madrid and Zayas 2007). After application,
pesticides may be transported to streams through either
T. Zha
School of Soil and Water Conservation, Beijing Forestry aerial deposition or runoff and soil drainage driven by wind
University, Beijing, China or rainfall events, respectively (Scholtz and Bidleman

123
Arch Environ Contam Toxicol (2013) 64:208–218 209

2007; Willis and McDowell 1982). Aerial deposition is according to hydrograph conditions represents the best
often minor during dry seasons or base flow periods (Ernst potential option for obtaining the complete range of con-
2007, 2009; Environment Canada 1982; Yao et al. 2006); centrations present in recipient waters. This study explored
transport of residues by runoff is much more important this hypothesis using two data sets concurrently collected
(Ernst 2007; Laabs et al. 2002; Stephenson and Solomon in a pesticide residue project (from 2003 to 2006) in the
2007). Although regular GS may collect more samples than Maritime region of Canada that used both GS and hydro-
required during dry seasons, this approach may in turn graph-based sampling.
result in fewer samples than required during wet seasons
and may not accurately reflect the hydrologic pattern of
pesticide contamination, a function of rainfall and water- Material and Methods
shed characteristics. GS is also ineffective because of the
so-called ‘‘first-wash’’ effect, meaning that the first rainfall Research Sites and Operating Dates
event or the initial phase of a rainfall event after applica-
tion may flush the greatest proportion of residues (Willis The monitored sites in this study (from 2003 to 2006)
and McDowell 1982). The nature of GS and its intrinsic include: (1) Little River Watershed (LRW, 47°020 to
inconsistency means that this type of sampling is difficult 47°200 N, 67°310 to 67°470 W), which is 380 km2 in size and
to conduct repetitively and collect representative samples under a mixture of agricultural (16.2 %), forest operation
to compare temporal trends (Laabs et al. 2002). This is (77 %), and residential (6.8 %) land uses; (2) Black Brook
compounded by the large volumes of water necessitated for Watershed (BBW, 47°050 to 47°090 N, 67°430 to 67°480 W),
some pesticide-residue analysis. which is 14.5 km2 in size and a subwatershed of LRW with
A more representative picture of water quality for risk agricultural land use constituting 65 % of the land base of
assessment may be obtained through a more comprehen- the watershed with the remainder either forested or under
sive approach that goes beyond simply increasing GS fre- urban and residential uses; and (3) subwatershed #9
quency. Automatic time-sequential sampling can be used to (sub#9) of the BBW, with 93 % of the total 0.8 km2 under
produce composite samples, but this does not necessarily agricultural production that had contour diversion terraces
capture hydrographic variations of stream flow over time and grassed waterway soil conservation systems installed
and may miss not only the maximum residue concentra- during 1996 to 200. The major crop in the study area was
tions but also the highest number of detectable pesticides potato in rotation with grain and hay, and the pesticides
associated with the peak flow in a storm runoff (Kimbe- monitored were associated with the major in-use com-
rough and Litke 1996). Passive-sampling has been pro- pounds at the time of the study (Chow et al. 2011); these
moted for its potential to become a reliable, robust, and are listed in Table 1.
cost-effective tool because it generates a time-integrated
average analyte concentration (Madrid and Zayas 2007). If Equipment Set-Up and Sampling
the water flows do not vary significantly, this approach can
be useful; however, because pesticides typically show In this research, two sampling systems were used: weekly
temporal variability coinciding with the hydrograph, this GS and hydrograph-based sampling. Hydrograph-based
approach is therefore limited in its ability to capture the full sampling was performed only during non-ice covered
range of pesticide residues. The accurate calculation of periods of the year (from late May to November for the
pesticide transport through runoff can only be obtained study region), whereas GS was performed year-round.
through the hydrography of pesticide concentration tracked
over the entire runoff event. With the development of Grab Sampling
computation technology, it is now possible to develop
process-based models of pesticide movement over runoff Grab samples were manually collected weekly from the
flow to obtain more accurate risk assessments of pesticide stream channel at each site in 2-L precleaned amber glass
contamination in the environment. Because the toxicity of bottles at 12 to 15 cm below the surface of the water at a
a given pesticide residue in stream water is governed by the point with maximum water mixing.
concentration and duration of the residue, it is necessary to
capture the concentration profile to obtain an accurate Hydrograph-based sampling
loading and assess any environmental risks. GS may not
able to provide such detailed information because of low Hydrograph-based sampling (Fig. 1) was automatically
sample frequency and, potentially, sampling at the wrong performed at constructed monitoring stations for each
time. Our hypothesis is that a continuous, hydrograph- rainfall event or base flow period. The sampling system
based sampling system that collects water samples consisted of a modified ISCO 2900 sampler (Teledyne

123
210 Arch Environ Contam Toxicol (2013) 64:208–218

Table 1 CCME guidelines for freshwater aquatic life, estimated safe levels of pesticides in surface water, and drinking water quality
Pesticide MDL (lg L-1)a Classb Chemical FAL guideline Estimated safe level Drinking water
family (lg L-1) guideline (lg L-1) guideline (lg L-1)

Azinphos-methyl 0.03/0.005 I OP 0.2b 0.023 20


Carbofuran 0.04/0.00024 I Carbamate 1.8 – 90
Chlorothalonil 0.02/0.00144 F Nitrile 0.18 – 70
Dimethoate 0.04/0.0066 I OP 6.2 – 20
Imidacloprid 0.02/0.0066 I Chloronicotinyl 0.2b 6.83 –
Linuron 0.06/0.00999 H Sulfonylurea 7 – –
b
Metalaxyl 0.03/0.00042 F Acylalmine 0.2 25003 500
Metribuzin 0.03/0.005 H Triazine 1 – 80
FAL freshwater aquatic life, OP organophosphates
a
DL number before ‘‘/’’ was used in ALET, whereas the number after ‘‘/’’ was used in NLET
b
No CCME guideline available. Estimated ‘‘safe’’ level at 0.2 of the lowest LC50 or 0.4 of the lowest chronic toxicity end point. If both types of
data were available, the lowest of these two numbers was used

maximum water mixing within the water column in the


stream through extended Teflon sucking tubing. The
modified sampler had a capability to collect a 2-L water
sample each time when activated by a CR10X data logger
(Campbell Scientific, Logan, UT, USA) triggered by the
stream stage height changes obtained by the stream stage
height-monitoring system and, totally, a maximum of 12
samples could be collected during a storm event or a
sample for every 72-hour period if there was no rainfall
during the period. The sampler was operated in flow mode,
and the sampling number was fed back to the data logger
for recording.
The stream stage height-monitoring system consisted of a
10-turn precision potentiometer operated using a float-
counter weight arrangement (Lakewood Systems, Edmon-
ton, Alberta, Canada). The stage height was monitored at a
Fig. 1 A water-level monitoring and hydrograph-based sampling 5-minute scanning intervals and recorded on an hourly basis
system for pesticide contamination study in the BBW, New Bruns- if the change in stage height was \ 2 mm compared with the
wick, Canada. Top system housing; Right stage height monitoring;
Bottom modified sampler with 2L bottles previous recorded value. Additional height data were
recorded if the change in stage height was C 2 mm (Chow
ISCO, NB, USA) and a flow-monitoring system. The flow- et al. 2011; Xing et al. 2012). The discharge rate was then
monitoring system was located in an instrument hut with a computed from the stage height using a predetermined, site
stilling well connected to the stream through a polyvinyl- specific rating curve. The rating curves for the BBW and
chloride pipe. sub#9 were first established based on a composite culvert
The standard ISCO 2900 automated water sampler in configuration-4-to-1 V-notch sharp-crested weir and rect-
12-bottle model was modified to accommodate 12 2-L angular-1-to-1 V-notch sharp-crested weir, respectively,
amber bottles by replacing the sample bottle tub with a and improved gradually through calibrations based on area
wood box (Fig. 1 [bottom left panel]) and installing a velocity using a current meter along a transect at 30-cm
custom-made sample-dispatching port ring on the top of intervals and 0.6 cm depth from the water surface (Xing et al.
the box and fitted to the central section of sampler. The 2012). The rating curve for the LRW was also based on the
dispatching ring had 12 ports lined along a ring, and each area velocity method measured at different stage heights. To
port led to a plastic tube that was connected to the rubber decrease power consumption and time required for compu-
lid of each sample bottle. The ports were numbered from 1 tation by the data logger, the conversion of stage height to
to 12 and located exactly at the point where the outlet of discharge rate was performed during post-data processing
the distributor arm pointed to during each sampling stop. using a laboratory computer (Xing et al. 2012). Due to the
The sampler in-take strainer was placed at a point with unavailability of real-time flow information and because of

123
Arch Environ Contam Toxicol (2013) 64:208–218 211

the elapsed time required for the sampler to collect 2 L of for precipitation amount and intensity (WS no. 5), which
sample, it was found that hydrograph-based sampling based were only monitored with heated (WS no. 8 [Rimco heated
on stage height was just as favorable as true flow proportional gauge; McVan Instrument, Mulgrave, Victoria, Australia])
sampling and provided acceptable data. The sampler was and nonheated (WS no. 5 [Texas Electronics, Dallas, TX])
activated by the data logger to take a sample at every 3- to tipping-bucket rain gauges. Because of the extremely low
5 cm change in stage height. These numbers were selected ambient temperature during the winter, an additional
based on the size of the drainage area and the most probable 200 W heater controlled by a thermostat (closed at 4 °C
rainfall amount and intensity and could be easily adjusted and open at 16 °C) was wrapped around the collection
based on the predicted storm from weather forecasts so that a funnel to accelerate snow melting before funneling it into
maximum of 12 samples could be collected during any major the tipping bucket. This modification was found to be
rainfall-runoff event. The reference stage height used for essential for the severe winter conditions in northwestern
sampling was readjusted to the new height automatically if New Brunswick, Canada. The sampling frequency was set
the difference between the current reading and the previous at 1-minute intervals during precipitation events to capture
sampling height was B-3 cm during the recessing phase of the temporal pattern of the rainfall event, whereas only
the hydrograph. In additional to triggering the water sampler hourly output was sent to storage during dry periods. Data
to collect water samples, the data logger also captured the were retrieved on a weekly basis by way of telephone line
sample number generated by the sampler and stored the using an answering modem in conjunction with the data
information with a time stamp. To synchronize the output logger.
frequency of the sampler to the CR10X logger, a 20 kX
resistor was required between the ‘‘pulse channel’’ and Laboratory Analysis
‘‘ground’’ when it was configured for high-frequency pulse
input from the sampler. With the slower operating frequency NLET Analysis Procedures in the NWRI
of the CR10 data logger, no resistor was needed. To mini-
mize pesticide adsorption, all tubing in the hydrograph- Samples collected before 2006 were spiked with 1.0 ml
based sampling system, except for the silicone tubing used 240 ppb carbaryl 13C and shaken well to ensure thorough
for the peristaltic pump, was replaced by Teflon. mixing. The samples were then stored (normally not [2 d)
All pesticide samples were preserved in the field with at 4 °C for further filtration and extraction. Whatman glass
5 ml methylene chloride. They were collected and put microfibre filters (0.45 lm; Whatman, Toronto, ON,
through a preprocessing procedure in the soil laboratory of Canada) were used to filter samples through a specialized
the AAFC research centre as described in Laboratory pressure-filtration system. The solids recovered from
Analysis. They were then packed on ice and shipped by filtration were funneled into a small bottle and stored at
courier to the National Laboratory for Environmental -20 °C. The aqueous samples were further extracted using
Testing (NLET), National Water Research Institute a solid phase extraction (SPE) manifold and Supelco 2-ml
(NWRI), Burlington, ON, Canada, for further analysis. reversible tubes with SPE frits and 500 mg ENVI-Carb C
Pesticide samples collected in 2006 were preserved in the cartridges (Supelco, Supelco Park, Bellefonte, PA, USA).
field with methylene chloride, packed on ice, and shipped SPEs containing pesticide residues were placed in self-
by courier for next-morning delivery to the Atlantic Lab- sealing polyethylene bags and frozen at -20 °C. The SPEs
oratory for Environmental Testing (ALET), Environmental and filtered solids were then shipped on ice overnight to the
Quality Laboratory, Environment Canada, Moncton, NB, NWRI laboratory for analysis. Cartridges were eluted by
Canada, where they were analyzed for pesticide concen- gravity in the reverse direction with 10 ml DCM and
tration. Because we did not have comparable GS data, the methanol (80:20). The eluents were then dried under
data collected in 2006 with the hydrograph-based sampling nitrogen and diluted in methanol for analysis by liquid
method were only used for sensitivity analysis of the chromatography with tandem mass spectrometry detection
hydrograph-based sampling method. (MS/MS; Sciex API 2000; Thornhill, ON, Canada) as
described in Chow et al. (2011). Detection limits for the
Weather Information methods used in this study are listed in Table 1.

Weather data (rainfall) used for this study were collected ALET Analysis Procedures in Environment Canada
from two automated weather stations, one located approx-
imately 200 m from the outlet of the BBW (WS no. 5) and The samples collected in 2006 were analyzed in the
the other one located approximately 5 km from the outlet Environment Canada Laboratory. Within 24 hours of
of the LRW at a permanent runoff-erosion study site (WS receipt, the 1-L water sample was filtered, using positive
no. 8). Both weather stations were run year-around, except pressure, through Whatman glass filters (GF)/C

123
212 Arch Environ Contam Toxicol (2013) 64:208–218

(N 14.2 mm; 1.2-lm pores) and GF/D (N 12.5 mm; 2.7 um Mean concentration differences between both methods
pores; Whatman, Fisher Canada, Oshawa, ON, Canada). were also apparent. Mean concentrations obtained through
The water fraction was then extracted with 3 9 50 mL GS were at least 50 % less using hydrograph-based sam-
methylene chloride (pesticide grade). The extract was pling (Table 2). Although these concentrations were well
concentrated to 5 mL and then passed through a gel-per- below drinking water-quality guidelines (Table 1) and
meation chromatography column (200 to 400 mesh; Bio- would not represent any risk to human health, the GS
Beads S-X3, Fisher) to remove coextractives. The extract method clearly underestimated the presence of pesticides in
was again concentrated, with the solvent exchanged to stream water.
toluene, to a final volume of 0.5 mL. The extract was The higher percentage of detections and higher con-
subsequently analyzed for pesticides using GS-MS centrations obtained through hydrograph-based sampling
(GC-5890/MS-5973; Agilent, Lexington, MA, USA) can be attributed to the better sampling design, i.e., addi-
operating in the selected ion monitoring mode. tional samples were collected when the stage height
increased or decreased by 3 cm compared with the stage
height of the previous sample. With this sampling
Results and Discussion approach, B12 samples could be collected for any single
storm event (e.g., July 21, 2006 [Fig. 2]). With hydro-
Detection Frequency and Concentration graph-based sampling, temporal variations in the stream
hydrograph during rainfall-runoff events were properly
Hydrograph-based sampling generally obtained higher delineated by the flow-controlled samples, and therefore a
detection rates of the listed pesticides (2 to 26 %) compared complete representative hydrograph of pesticide concen-
with GS (1 % [Table 2]). When the method detection limit trations with event progress could also be accurately cap-
(MDL) (Table 1) threshold was used on 3 years of col- tured (Fig. 2). In contrast, GS provided only a general
lected samples in the BBW (2003 to 2005) using GS, no representation of the stream water hydrograph through
cases were detected that exceeded the MDL for metribuzin, regularly collected samples. The probability of capturing
metalaxyl, and azinphons-methyl, whereas only one case pesticides had a 1-week resolution, and runoff sampling
each was detected that exceeded the MDL for imidacloprid occurred mainly by chance because our results clearly
and linuron. demonstrate that the primary route of pesticides entering
Although the maximum detection rate among pesticides these streams is by way of rainfall runoff events (e.g.,
was \27 % for both approaches, differences between the Fig. 3).
two methods were apparent. There were no detections of
pesticides under normal (base) flow conditions regardless Maximum Concentration
of sampling approach. However, with samples during and
immediately after major rainfall events, it was found that As part of sampling comparison, data collected for imi-
there were some detections of pesticides in excess of the dacloprid during the growing season of 2005 are provided
MDL but below available water-quality guidelines for in Fig. 3. It was found that both methods generally cap-
aquatic life for both methods. Careful examination at the tured the base flow variations of imidacloprid; however, the
individual detection rates of pesticide obtained using hyd- concentrations obtained through the hydrograph-based
rograph-based sampling in the BBW indicated detection method showed greater concentrations (by 1 to 3 orders of
rates from 30 % for linuron to 0 % for metribuzin (average magnitude) than those obtained by way of GS. The one
9 %) for 2005. exception to this was when GS occurred during a rainfall

Table 2 Detection rates higher than MDL and corresponding concentrations obtained through HBS and GS from 2003 to 2005 at the outlet of
BBW
Sampling No. of Parameters Imidacloprid Metribuzin Metalaxyl Linuron Azinphos-methyl
method samples

HBS 163 % [MDL 7.36 2.45 3.68 26.38 9.20


Mean concentration (lg L-1) 0.46 ± 0.05 0.09 ± 0.02 0.69 ± 0.02 0.25 ± 0.03 1.88 ± 0.7
GS 105 % [MDL 1.0 0 0 1.0 0
-1
Mean conentration (lg L ) 0.22 ± 0.00 – – 0.09 ± 0.00 –
Mean concentration (±SE) was calculated from measurements greater than MDL
HBS hydrograph-based sampling

123
Arch Environ Contam Toxicol (2013) 64:208–218 213

Fig. 2 An example of 80
Stage HT (cm)
hydrograph-based sampling Stage HT change (cm)
data. Data were collected at the Sample #
outlet of the Black Brook Rainfall Acc. (mm)
Watershed, NB, Canada. Rainfall amount (mm)
Sample # series number of the 60 in 30 minutes
samples collected, Rainfall Acc
accumulative rainfall amount

Measurement
since the start of the rainfall
event, Stage HT stage height
40

20

0
13:00:00 14:00:00 15:00:00 16:00:00 17:00:00
Time (July 21, 2006)

Fig. 3 Monitored 4
Imidacloprid (ng L-1, log transformed)

concentrations of the pesticide Daily rain amount (mm)(log transformed)


imidacloprid measured during Grab sampling
2005 at the outlet of the BBW 3 Hydrograph-based sampling
using both sampling
approaches. RF rainfall
2

1 10 mm RF

0 1 mm RF

-1
120 140 160 180 200 220 240 260 280 300 320
May 20 Day of year 2005 Oct 27

event as indicated on day 142 (Fig. 3). It should also be residues. Grab sampling was weekly during this period;
noted that [10 mm daily rainfall could make relatively however, for the hydrograph-based sampling method, data
large differences in terms of maximum concentrations. It is were not continuous but event-triggered. In such cases, if
possible that grab samples might have mostly been col- there was no rainfall and thus no data for hydrograph-based
lected either before or after major rainfall events when sampling, the data from GS were used for the calculations
concentrations of pesticides in the stream were very low. In of the base flow period. For rainfall events, calculating
contrast, the hydrograph-based sampling method sampled accumulative pesticide from hydrograph-based sampling
mostly during rainfall events and could capture both ‘‘first was expressed as the integration of real measurements of
wash’’ and the subsequent dilution effect of rainfall on water discharge multiplied by measured pesticide concen-
pesticide residues running into stream water. tration. The results showed an average of 43 % underesti-
mation using the GS method (Table 3). However, the
Accumulated Pesticide Transportation difference will be pesticide-dependent. For frequently
detected pesticides, the differences between the two
Data collected using both methods from the BBW during methods were relatively lower and insignificant (p [ 0.05
May to October 2005 were used to derive a cumulative for imidacloprid, metalaxyl, and azinphos-methyl
loading (concentration 9 flow discharge) of pesticide [Table 3]), whereas the differences were greater and

123
214 Arch Environ Contam Toxicol (2013) 64:208–218

Table 3 Comparison of accumulated pesticide amount detected in BBW during May through Oct 2005 by way of two sampling methods
Sampling method Imidacloprid (kg) Metribuzin* (kg) Metalaxyl (kg) Linuron* (kg) Azinphos Methyl (kg)

HBS 41.66 1.61 2.09 56.17 0.56


GS 36.05 0.22 2.00 12.50 0.43
% 13.48 86.22 4.23 77.74 23.11
p value for ANOVA 0.411 0.027 0.922 0.021 0.243
% = (HBS-GS)/GS*100
HBS hydrograph-based sampling, ANOVA analysis of variance
* Significant difference (p \ 0.05) between sampling approaches
Weather variables
Fig. 4 Relationships between
50
(a) rainfall characteristics,
40 a
(b) hydrographs, and Accumulative. Rainfall (mm)
30 since day 180
(c) dimethoate concentrations 20 30 minute rain intensity (mm)
detected using hydrograph- 10
based sampling from three 0
watersheds of different size.
Watershed sizes are 76.7, 1450, 180.0 180.5 181.0 181.5 182.0 182.5 183.0
and 38,000 for sub#9, BBW,
River stage height (cm)

and LRW, respectively


100
Sub#9
80
b BBW
LRW
60

40

20

0
180.0 180.5 181.0 181.5 182.0 182.5 183.0
Dimethoate (µg L-1)

.35
.30 c Sub#9
.25 BBW
LRW
.20
.15
.10
.05
0.00

180.0 180.5 181.0 181.5 182.0 182.5 183.0


Julian Day of 2006

significant (p \ 0.05, Table 3) for pesticides associated (solid line), BBW (dotted line), and LRW (dashed line). It
with events. This indicates the weakness inherent in GS in should also be noted that the peak time of stage heights
that it captures pesticide residue by chance. measured at BBW were generally behind those of sub#9 by
1 to 2 hours, and those of LRW were behind by an addi-
Spatial Sensitivity tional 1 to 2 h. For instance, the first peak for sub#9
occurred at 180.747 day compared with 180.795 day for
Figure 4 shows the spatial and temporal changes in rainfall, BBW and 180.875 day for LRW. The time lag between
river stage height, and dimethoate concentrations measured sub#9 and BBW was 69 minutes, whereas it was 115
at the outlets of sub#9, BBW, and LRW during a series of minutes between BBW and LRW. This difference is due to
rain storm runoff events from June 28 to 30, 2006. As watershed characteristics including topography, geology,
shown in Figure 4a, the accumulative rainfall since day and land use of the watersheds and illustrates the ability of
180 was 41.4 mm with corresponding 30-min maximum the hydrograph-based sampling to measure hydrological
rainfall intensities of 16.3, 12.2, 14.2, 5.6, and 6.6 mm h-1, differences between watersheds.
respectively, for the five rainfall events. Accordingly, the Similarly, Figure 4c shows the dimethoate concentra-
stage heights (Fig. 4b) recorded by hydrograph-based tions of samples collected during these events. Dimethoate
sampling clearly reflected each event at the outlets of sub#9 was not detected at the outlet of any of the monitored

123
Arch Environ Contam Toxicol (2013) 64:208–218 215

watersheds during base flow or before the rain event at Other factors, such as land-management patterns, crop
approximately 17:00 (Julian day 180.7). With event pro- cover type, cropping intensity, soil drainage condition, and
gression, concentrations varied with different watershed topographic configuration, may also be responsible for
sizes, rainfall intensity, and rainfall duration. Dimethoate spatial variations in pesticide residues in surface water
concentrations measured from the LRW showed very little among different watersheds. Land fragmentation, grassed
variation and no observed stage height changes (Fig. 4c), waterways, and riparian buffer strips may prevent pesticide
whereas concentrations from the BBW displayed only one residues from reaching streams as seen in sub#9. All of
peak response associated with the largest stage height these processes potentially decouple the linear relationship
change. However, dimethoate concentrations from sub#9 between rainfall intensity, duration, and pesticide transport
had several peak periods corresponding to the peaks of amount and patterns. However, the degree of the decouple
stage heights. These observations may reflect a relationship may varied with watershed characteristics. Therefore,
between crop intensity, pesticide-application rates, and determining the appropriate sample collection time is
resultant pesticide concentrations detected between critical for capturing the ‘‘first wash’’ phenomenon of
watersheds. pesticide residue transportation and, in particular, the
Pesticide concentrations in runoff from the nonagricul- maximum concentration of pesticide residues in stream
tural portion of the watershed were considerably less than water during a rainfall event for different watersheds.
those of the agricultural areas (Fig. 4). Land-use propor-
tions for agricultural production in LRW, BBW and sub#9 Temporal Sensitivity
were 16, 65, and 93 %, respectively, and may be the pri-
mary factor for lower concentrations at the outlets in the To demonstrate the sensitivity of the hydrograph-based
LRW than in the BBW. Furthermore, sub#9 was the sampling method in capturing pesticide concentrations and
smallest watershed, and its outlet was the closest to the stream flow resulting from different rainstorms, data col-
fields/pesticide source and less impacted by any attenuation lected from three rainfall events in 2006 (June 27, July 21,
in runoff path and therefore incurred a less net loss of and August 15) in the BBW are provided in Fig. 5. Both
pesticide residues transported by runoff. stage height and total suspended sediment concentrations

Fig. 5 Stage height, total 40


sediment concentration, and
a Rainfall (14.22 mm) Chlorothalonil ( g L )
-1

June 27 Metribuzin ( g L )
-1

pesticide measurements 30 -1
Linuron ( g L )
(chlorothalonil, metribuzin, and Stage height (x10cm)
-1
20 TSSC (g L )
linuron) obtained at the outlet of
sub#9 of the BBW during three 10
rainfall events (a–c) during
2006. Stage height stream stage 0
height recorded by data logger, 03:30 04:00 04:30 05:00 05:30 06:00 06:30
TSSC total suspended sediment Time (June 27 2006)
concentration
2.5
b Rainfall (11.68 mm)
July 21
-1
Chlorothalonil (mg L )
2.0 -1
Metribuzin (mg L )
Measurements

-1
1.5 Linuron (mg L )
Stage height (x10cm)
-1
1.0 TSSC (g L )

.5

0.0
13:30 14:00 14:30 15:00 15:30 16:00
Time (July 21 2006)

4
c Rainfall (9.4mm) Chlorothalonil (x100 g L )
-1
August 15 -1
3 Metribuzin ( g L )
-1
Linuron ( g L )
Stage height (x10cm)
2 -1
TSSC (g L )

0
16:30 17:00 17:30 18:00 18:30 19:00 19:30 20:00 20:30
Time (August 15 2006)

123
216 Arch Environ Contam Toxicol (2013) 64:208–218

displayed polynomial curve patterns skewed significantly measurement date. Furthermore, in this region, chlorot-
to the left, indicating the increasing configuration of the halonil was mainly used in July and early August for
hyetograph, thus reflecting rainfall characteristics. It must controlling late potato blight. The variations from June to
be noted that the impacts of the phenologic development of August clearly demonstrate the application pattern of the
crop canopy on runoff and pesticide transport were also pesticide in this region. This hydrographic temporal pattern
inherent in the data set because the rainfall events occurred shown by hydrograph-based sampling could not be cap-
in different crop-development phases. From these data, it is tured by GS because it has a coarse resolution (±1 week).
interesting to see that the total suspended sediments The known behavior of metribuzin is consistent with the
reached their peak fluxes approximately half an hour later data obtained from June 27 to August 15 2006 (Fig. 5a–c).
than the corresponding stage height (Fig. 5) and were less Because it is a pre-emergence herbicide, metribuzin was
related to the measurement dates (rainfall characteristics). applied only in late May or early June. Metribuzin is mainly
Pesticide concentrations throughout the rainfall events removed from soil by microbial degradation with leaching to
showed similar patterns to the suspended sediments groundwater. Its half-life in soil ranges between 2.5 and
(Fig. 5); however, they were pesticide-dependent. The 4 months, but it was found to be only approximately 7 days
maximum concentrations of chlorothalonil had a bell- in pond water (see http://dsp-psd.pwgsc.gc.ca/collection/
shaped curve, except for August 15, during most of the H48-10-1-58-1989E.pdf). Given the previously mentioned
rainfall events (Fig. 5c). Pesticide concentration peaks characteristics, the concentration gradients from approxi-
mostly corresponded with peaks in total suspended sedi- mately 5 to 10 lg L-1 in June to very close to zero in August
ments. Then the concentrations of the pesticides showed (Fig. 5a–c) was evidence of the accuracy and sensitivity of
either a decay curve pattern with decrease in rainfall the hydrograph-based sampling approach in capturing long-
amount, a demonstration of ‘‘first wash’’ effect of rainfall term temporal variation of pesticide residues.
(as in Fig. 5c), or normal distributions (Fig. 5a, b), which Linuron is an herbicide and did not show much vari-
may be attributed to pesticide characteristics, such as water ability with different rainfalls except for the rainfall on
solubility and soil absorption. The ‘‘first wash’’ effect is June 27. It is well known that in the environment, linuron is
mostly affected by the intensity and duration of the rainfall. moderately persistent in soils and is broken down, mainly
As shown in Fig. 5, concentrations varied greatly from by microbes, into a much less active metabolite with a half-
June (\35 lg L-1) to July (500 to 1000 lg L-1), and life of 75 days (see http://pmep.cce.cornell.edu/profiles/
August (\100 lg L-1), possibly caused by the amount of extoxnet/haloxyfop-methylparathion/linuron-ext.html).
pesticide applied, number elapsed days since the dates of Furthermore, linuron is only slightly to moderately soluble
pesticide application, and rainfall characteristics at each and is not readily broken down in water (Kalkhoff et al.

Table 4 Statistical
Sampling dates Imidacloprid (lg L-1)* Linuron (lg L-1)*
comparisons of the
concentrations of two pesticides HBS GS HBS GS
measured during July 27 to
September 14, 2004, by HBS Julian day
and GS 208 0.4196 0.0042 0.3995 0.0107
216 0.0480 0.0060 0.1466 0.0051
222 0.0097 0.0063 0.0074 0.0029
229 0.0072 0.0025 0.0068 0.0009
236 0.0341 0.0019 0.0271 0.0007
243 0.0023 0.0023 0.0185 0.0009
251 0.0070 0.0025 0.0555 0.0008
257 0.0044 0.0035 0.0327 0.0054
Period mean 0.0665 0.0037 0.0868 0.0034
SE 0.0508 0.0006 0.0475 0.0012
Range of mean 0.4174 0.0044 0.3926 0.0100
HBS hydrograph-based
Minimum 0.0023 0.0019 0.0068 0.0007
sampling
Maximum 0.4196 0.0063 0.3995 0.0107
* Significant differences
(p \ 0.05) in Student t tests Range of real measurements 0.6202 0.0044 0.6844 0.0099
were found between two Relative difference (%) 94.5 96.0
sampling approaches for a Student t test p 0.028 \0.001
pesticide (log-transformed data)

123
Arch Environ Contam Toxicol (2013) 64:208–218 217

2003). Therefore, the concentration decrease from 5 lg to 3 orders of magnitudes. Hydrograph-based sampling
L-1 on June 27 (Fig. 5a) to \0.5 lg L-1 on August 15 described the spatial and temporal variability of pesticides
(Fig. 5c) showed a time-related pattern and suggested that in streams of various-sized watersheds more accurately
the hydrograph-based sampling method reasonably cap- than did GS.
tured the impacts of pesticide characteristics. Statistical analysis displayed significant differences in
the means of pesticide measurements obtained with the two
Statistical Comparisons methods. GS provided insufficient resolution to make
conclusions regarding the mobility, and, ultimately the risk
A data segment from day 208 to 257 in 2004 from BBW of pesticides in the environment.
was used to carry out statistic analysis as a demonstration. Hydrograph-based sampling can be used for pesticide
The result (Table 4) showed that the two sampling contamination research and is a more precise tool for
approaches were significantly different in terms of daily gaining insight into contamination patterns and collecting
mean, range, maximum, and minimum concentrations of detailed information for modeling the environmental
the pesticides imidacloprid and linuron. Student t test dis- impacts of pesticides. Hydrograph-based sampling showed
played a significant difference between the two methods a profile-style resolution with the progress of rainfall and
for imidaloprid (p = 0.014) and linuron (p = 0.028). GS displayed a reasonable pattern of pesticide-residue con-
underestimated concentration means by 95 and 96 % for centrations in stream water based on a maximum of 12
imidacloprid and linuron, respectively. Maximum concen- water samples for any given storm. However, the limited
trations of imidacloprid and linuron detected by hydro- number of samples for which the sampler was designed
graph-based sampling were two to three orders of may be inconvenient for consecutive rainfall events and
magnitude greater than those detected by GS, respectively. thus create more data gaps. To adapt the method to more
This result also suggests that average- or median-based complicated rainfall patterns, composite sampling using the
parameters may not be appropriate to describe pesticide hydrograph-based-flow proportional sampling scheme
contamination in stream water for risk assessment purposes (composite of a number of samples into one sample based
because of temporal-distribution patterns of pesticide on a predetermined amount of discharge) may be a superior
movement. approach to capturing the concentration per rainfall event
There is some concern that that hydrograph-based and decreasing sample number. This will make obtaining
sampling overestimates pesticide presence in stream water mean concentrations over an entire year for a long-term
because the method obtains pesticide concentrations during statistical analysis possible and meaningful.
high flow. In this study, we determined that pesticide
transport was mostly associated with runoff flow (verifying Acknowledgments The authors acknowledge the work of John
Monteith, Lionel Stevens, Sylvie Lavoie, Suzanne Batchelor, and Yu
the ‘‘first-wash’’ phenomenon) and that base-flow detec- Yang for their contributions in maintaining the weirs, sample col-
tions were generally zero or below detection limits. lection, and analysis. This research project was financially supported
Therefore, the exact ‘‘washing-time’’ that pesticides flow by Watershed Evaluation of Beneficial Management Practices
into streams will be determined by the flow rate and rainfall research project of AAFC, Canada.
amount, which is normally unknown before a rainfall.
From this aspect, hydrograph-based sampling provides a
References
more accurate approach because this method captures this
all-important transport window. Chow L, Xing Z, Benoy G, Rees HW, Meng F-R, Jiang YF et al
(2011) Hydrology and water quality across gradients of agricul-
tural intensity in the Little River Watershed area, New Bruns-
Conclusion wick, Canada. J Soil Water Conserv 66(1):71–84
Day KE (1990) Pesticide residues in freshwater and marine
zooplankton: a review. Environ Pollut 67:205–222
The hydrograph-based sampling method aptly character- Environment Canada (1982) Canadian Climate Normals 1951–1980.
ized pesticide pollution in stream water. It detected more Temperature and precipitation Atlantic Provinces. Atmospheric
pesticide residues than GS by 20 to 30 %. The hydrograph- Environment Service, Environment Canada, Ottawa, Canada
Ernst W (2007) Assessing pesticide fate and biological effects from
based sampling method better obtained the maximum potato agriculture. Environmental Canada PSF Project Annual
concentration of pesticide residues in stream water, Report
whereas GS often missed peak concentrations. With GS, it Ernst B (2009) Surveillance of pesticide residues in surface water,
was possible to underestimate the impacts of pesticide sediment and groundwater in agricultural regions of Prince
Edward Island, New Brunswick and Nova Scotia. Environmental
residues on stream water quality, generally underestimating Canada PSF Project Annual Report
average pesticide concentration by B50 %, whereas the Ernst B, Mickle RE, Howatt S, Pitbaldo R, Nichols I, Rowlandson T,
underestimation of maximum concentrations could reach 1 et al. (2009) The development of meteorological standards for

123
218 Arch Environ Contam Toxicol (2013) 64:208–218

pesticide application in Canada. National Agri-Environmental Brunswick, Canada. Northeast Agricultural/Biological Engineer-
Standards Initiative Synthesis Report No 10. Environment ing Conference. American Society of Agricultural Engineers.
Canada Gatineau, Quebec, Canada The Pennsylvania State University
Harris ML, Wilson LK, Elliott JE, Bishop CA, Tomlin AD, Henning Peruzzo PJ, Porta AA, Ronco AE (2008) Levels of glyphosate in
KV (2000) Transfer of DDT and metabolites from fruit orchard surface waters, sediments, and soils associated with direct
soils to American robins (Turdus migratorius) twenty years after sowing soybean cultivation in the north pampasic region of
agricultural use of DDT in Canada. Arch Environ Contam Argentina. Environ Pollut 156:61–66
Toxicol 39:205–220 Scholtz MT, Bidleman TF (2007) Modelling of the long-term fate of
Kalkhoff SJ, Lee KE, Porter SD, Terrio PJ, Thurman EM (2003) pesticide residues in agricultural soils and their surface exchange
Herbicides and herbicide degradation products in upper midwest with the atmosphere: Part II. Projected long-term fate of
agricultural streams during August base-flow conditions. J Envi- pesticide residues. Sci Total Environ 377:61–80
ron Qual 32:1025–1035 Schulz R, Hauschild M, Ebeling M, Nanko-Drees J, Wogram J, Liess
Kanazawa J (1989) Relationship between the soil sorption constants M (1998) A qualitative field method for monitoring pesticides in
for pesticides and their physiochemical properties. Environ the edge-of-field runoff. Chemosphere 36:3071–3082
Toxicol Chem 8:477–484 Smith MC, Thomas DL, Bottcher AB, Campbell KL (1990)
Kimberough RA, Litke DW (1996) Pesticides in stream draining Measurement of pesticides transport to shallow ground water.
agricultural and urban areas in Colorado. Environ Sci Technol Trans ASAE 33:1573–1582
30:908–916 Stephenson GR, Solomon KR (2007) Pesticides and the environment.
Kolpin DW, Thurman EM, Lee EA, Meyer MT, Furlong ET, Canadian Network of Toxicology Centres Press, Guelph
Glassmeyer ST (2006) Urban contributions of glyphosate and its Williams RJ, Brooke DN, Matthiessen P, Mills M, Turnbull A,
degradate AMPA to streams in the United States. Sci Total Harrison RM (1995) Pesticide transport to surface waters within
Environ 354:191–197 an agricultural catchment. J Inst Water Environ Manage 9:72–81
Laabs V, Amelung W, Pinto AA, Wantzen M, da Silva CJ, Zech W Willis GH, McDowell LL (1982) Review: Pesticides in agricultural
(2002) Pesticides in surface water, sediment, and rainfall of the runoff and their effects on downstream water quality. Environ
northeastern Pantanal Basin, Brazil. J Environ Qual Toxicol Chem 1:267–279
31:1636–1648 Xing Z, Chow L, Cook A, Benoy G, Rees HW, Ernst B et al (2012)
Lakshminarayana JSS, O’Neill HJ, Jonnavithula SD, Leger DA, Pesticide application and detection in nine agricultural water-
Milburn PH (1992) Impact of atrazine-bearing agricultural tile sheds and their river systems in the maritime region of Canada.
drainage discharge on planktonic drift of a natural stream. Arch Environ Contam Toxicol 63(4):471–483
Environ Pollut 76:201–210 Yao Y, Tuduri L, Harner T, Blanchard P, Waite D, Poissant L et al
Madrid Y, Zayas ZP (2007) Water sampling: Traditional methods and (2006) Spatial and temporal distribution of pesticide air
new approaches in water sampling strategy. Trends Anal Chem concentrations in Canadian agricultural regions. Atmos Environ
26(4):293–299 40:4339–4351
Milburn P, Gartley C, Richards J, O’Neill H (1990) Effects of potato
production on groundwater quality: observations in New

123

S-ar putea să vă placă și