|
Next, the concentration of the pesticide in the environ-
<br />mental sample (zero if reported less than or estimated
<br />below the MRL) was subtracted from the concentra-
<br />tion in the spiked sample. Finally, this result was
<br />multiplied by 100 and then divided by the expected or
<br />theoretical concentration of the spiked sample to
<br />determine the pesticide recovery percentage.
<br />Pesticide surrogates were used to assess analyt-
<br />ical recovery and precision for each analysis. Surro-
<br />gate solutions containing known concentrations of
<br />organic compounds were added in the laboratory to
<br />each of the 155 environmental and quality-assurance
<br />samples prior to analysis. The surrogates were not
<br />expected to be found in the environmental samples but
<br />were expected to have similar chemical properties as
<br />the pesticides of interest. The surrogates diazinon-d1o~
<br />terbuthylazine, and alpha-HCH-d6 were added to the
<br />samples analyzed by GC/MS, and the surrogate
<br />BDMC was added to the samples analyzed by HPLC.
<br />Surrogates were reported as percent surrogate recov-
<br />ered and analyzed by mean surrogate recovery.
<br />No pesticides were detected in the seven field-
<br />blank samples. For four of the blank samples, the sites
<br />sampled just prior to the blank processing had various
<br />pesticides detected above the MRLs. Field-cleaning
<br />procedures were, therefore, effective in preventing
<br />contamination from one sample to another during
<br />equipment use. The potential for contamination of
<br />samples during sample collection, processing,
<br />cleaning, shipping, and analysis procedures was
<br />minimal.
<br />As a result of the replicate sampling, 88 repli-
<br />cate groups for 20 pesticides were studied (table 6), a
<br />replicate group here having at least one detected
<br />concentration equal to or greater than the MRL for
<br />the respective pesticide. Relative percent differences
<br />were determined for these replicate groups, and the
<br />differences generally were small (table 6). Differences
<br />ranged from 0.0 to 100 percent, and most were less
<br />than 20 percent. The only pesticide with a relative
<br />percent difference above 41 percent was 2,4-D. As
<br />such, variability in concentrations for the pesticide
<br />sampling generally was low. Forty-seven of the
<br />88 replicate groups consisted of a single pair of repli-
<br />cates, and 18 of these had a relative percent difference
<br />of 0.0 percent. Of the 41 replicate groups with
<br />multiple pairs of replicates, 22 contained at least one
<br />pair with no relative percent difference. Of the
<br />88 replicate groups studied, only 7 contained concen-
<br />trations that were reported both above and below the
<br />MRL for the respective pesticides and, thus, did not
<br />have a relative percent difference computed. Except
<br />for bromoxynil, these concentrations were less than
<br />0.008 µg/L. Overall, there was good consistency in
<br />identifying the targeted pesticides, and only at very
<br />low concentrations were pesticides apt to have concen-
<br />trations detected above and below the detection limit
<br />for the same replicate group.
<br />Mean pesticide recoveries in the 30 samples
<br />spiked in the field for analysis by GC/MS ranged
<br />from 34 to 227 percent with a median of 98 percent
<br />(table 7). Deethylatrazine, p,p'-DDE, disulfoton,
<br />cis-permethrin, and phorate all had mean recoveries
<br />below 70 percent, with deethylatrazine and
<br />cis-permethrin having the lowest recoveries, below
<br />40 percent. Azinphos-methyl, carbaryl, and carbofuran
<br />had high and variable recovery percentages. During
<br />testing of the GC/MS analytical method, Zaugg and
<br />others (1995) found azinphos-methyl, carbaryl,
<br />carbofuran, deethylatrazine, and terbacil to have low
<br />or highly variable spike recoveries. Because of this,
<br />analytical results for these five pesticides are reported
<br />by the NWQL as estimated (E) concentrations. There
<br />is increased uncertainty in analytical precision for
<br />estimated concentrations, but there is no increase in
<br />uncertainty of analytical detection (Rinella and Janet,
<br />1998).
<br />Mean recoveries for the spiked samples
<br />analyzed by the HPLC method were lower than those
<br />for the GC/MS method; the recoveries ranged from 13
<br />to 90 percent, and the median was 77 percent (table 8).
<br />Low and/or variable recoveries were reported for aldi-
<br />carb, aldicarb sulfone, carbofuran, chlorothalonil,
<br />clopyralid, dicamba, dichlobenil, methomyl, oryzalin,
<br />and picloram. Based on low or variable recovery
<br />performance, the NWQL reports concentrations of
<br />aldicarb, aldicarb sulfone, aldicarb sulfoxide,
<br />chloramben, chlorothalonil, dichlobenil, and DNOC as
<br />estimated values (Werner and others, 1996; National
<br />Water Quality Laboratory, 1998). As with the esti-
<br />matedconcentrations for the GC/MS analysis method,
<br />the analytical results for the pesticides with estimated
<br />concentrations as determined by the HPLC method are
<br />reliable detections with greater than average uncer-
<br />tainty for numerical precision.
<br />As mentioned previously, carbaryl and carbo-
<br />furan were analyzed by both the GC/MS and the
<br />HPLC methods. The mean recoveries for carbaryl for
<br />the two methods were 196 and 75 percent, respec-
<br />tively, whereas the mean recoveries for carbofuran
<br />14 Pesticides in Surface Waters of the Upper Colorado River Basin, Colorado, 1996-98
<br />
|