|
were 227 and 90 percent, respectively (tables 7 and 8).
<br />Because the recovery rates for the HPLC method were
<br />closer to 100 percent than the recovery rates for the
<br />GC/MS method, only the carbaryl and carbofuran
<br />occurrences and concentrations resulting from the
<br />HPLC method will be discussed in later sections of the
<br />report. With recovery rates below 100 percent, the
<br />carbaryl and carbofuran results from the HPLC
<br />method may be biased slightly low; however, the
<br />results from the GC/MS method would have a very
<br />high bias because of the recovery rates being around
<br />200 percent.
<br />For the pesticides with low recovery percent-
<br />ages, the detection frequencies and concentration
<br />ranges may be biased low. False negatives occur when
<br />a pesticide is present in a sample at a concentration
<br />greater than the MRL but is not detected because of
<br />analytical problems. In the UCOL study, for example,
<br />the mean recovery for dicamba was 58 percent (table
<br />8). Because of this low recovery, dicamba may not
<br />have been detected in a sample even though it was
<br />actually present above the MRL. The probability of a
<br />false positive (the detection of a pesticide at a concen-
<br />tration greater than the MRL when the actual concen-
<br />tration isbelow the reporting limit) is much lower than
<br />a false negative when the data are censored at the
<br />MRL (National Water Quality Laboratory, 1994).
<br />Mean recoveries for the pesticide surrogates
<br />terbuthylazine, diazinon-C~tp, alpha-HCH-d~, and
<br />BDMC were 112, 100, 99.9, and 84.0 percent, respec-
<br />tively (table 9). As with the spiked samples, the mean
<br />surrogate recovery for the BDMC samples analyzed
<br />by HPLC was lower than the mean recoveries for the
<br />terbuthylazine, diazinon-dip, alpha-HCH-d6 samples
<br />analyzed by GC/MS. Recoveries for all samples but
<br />one ranged between 64 and 141 percent, and most
<br />Table 9. Mean and standard deviation of pesticide
<br />surrogate recovery
<br />
<br />Pesticide surrogate Mean Standard deviation
<br />recovery
<br /> (percent)
<br />(percent)
<br />Gas Chromatography/Mass Spectrometry (GC/MS) analysis
<br />alpha-HCH-d~ 99.9 13
<br />diazinon-d~ p 100 1 1
<br />terbuthylazine 112 ]0
<br />High-Performance L iquid Chromatography (HPLC) analysis
<br />BDMC 84.0 11
<br />recoveries fell between 80 and 120 percent. A
<br />recovery of 0 percent was reported for one BDMC
<br />sample; this surrogate may have accidentally been
<br />omitted from the sample. These results of the surro-
<br />gate recoveries show that no significant bias occurred
<br />in the pesticide data, and variabilities in recovery, as
<br />shown by the standard deviations in table 9, were
<br />similar for the four surrogates.
<br />HYDROLOGIC CONDITIONS DURING
<br />SAMPLING
<br />Streamflow and runoff were higher than
<br />normal in the study unit during water year (WY) 1997
<br />(October 1996 through September 1997). The annual
<br />mean Streamflow and total runoff for the Colorado
<br />River near the Colorado-Utah State Line (encom-
<br />passing Streamflow from the entire study unit) for
<br />WY 1997 were 9,826 ft3/s and 7,114,000 acre-ft,
<br />respectively (U.S. Geological Survey, 1998b). In
<br />comparison, the long-term (WY 1970-97) mean
<br />annual streamflow and runoff amounts were
<br />6,847 ft3/s and 4,960,000 acre-ft, respectively. The
<br />annual mean streamflow and runoff amounts for the
<br />Gunnison River near Grand Junction for WY 1997
<br />also were above the long-term (WY 1970-97) aver-
<br />ages for this site. For calender year 1997, precipitation
<br />was 35 percent above normal at Grand Junction,
<br />mainly in April, May, August, and September. At
<br />Montrose, precipitation in 1997 was 21 percent above
<br />normal, mainly in April, May, June, and September
<br />(National Oceanic and Atmospheric Administration,
<br />1998a).
<br />Streamflow for the Colorado River near State
<br />Line and Gore Creek at mouth was dependent on
<br />snowmelt runoff; streamflow was generally highest in
<br />June during snowmelt (fig. 3). In contrast, streamflow
<br />in Reed Wash was caused primarily by irrigation and
<br />irrigation-return flow, with high streamflow from April
<br />into November. Streamflow for Dry Creek near
<br />Begonia Road was primarily dependent on irrigation
<br />and irrigation-return flow but also included snowmelt
<br />runoff from the Uncompahgre Plateau. For all four
<br />fixed-station sites, water-quality samples were
<br />collected over a wide range of streamflow conditions
<br />(fig. 3). In figure 3, samples were positioned based on
<br />instantaneous streamflow at the time of collection and,
<br />thus, may not plot on the line representing daily mean
<br />streamflow.
<br />HYDROLOGIC CONDITIONS DURING SAMPLING 21
<br />
|