When five of these samples were dried, ground,
One of the 25 ratios was excluded because it ex-
and split prior to analysis by both labs, the 20
ceeded 3.00. The 95% confidence limits on the
pairs of results for the same analytes yielded only
geometric means bracketed the expected mean of
one ratio outside 0.254.00. In fact, the range of
1.00 very nicely in both cases. However, the low
the other 19 ratios was 0.441.16; the geometric
number of degrees of freedom in the standard
mean was 0.79, 95% confidence limits for the mean
deviation estimates caused the 99% tolerance in-
were 0.700.89, and 99% tolerance intervals were
tervals to be wider than desired, especially for
0.391.61. It is unclear why a significant bias ex-
1,2-dichloroethene (0.392.42). If one uses the ac-
isted between the laboratories when none was
ceptability criterion suggested previously for met-
present for the samples split in the normal way.
als (0.502.00), all 24 ratios are within these lim-
However, it is most important that the standard
its. Although there are insufficient data to permit
deviation of the ratios for the dried and ground
evaluations of the other VOCs, there is nothing to
samples was considerably less than for the origi-
indicate that this range would be unsuitable for
nal splits (significant at the 99% confidence level).
them. On the contrary, the fact that these data
Interestingly, if the bias between laboratories is
arise from a single multianalyte procedure sug-
corrected to a mean ratio of 1.00, the 99% toler-
gests that the precision should be quite similar
ance intervals would become 0.492.04. This pro-
for each of the analytes. This differs from the situ-
vides considerable evidence for the benefit of
ation with metals where several methods are em-
sample drying and grinding prior to splitting.
ployed. Therefore, we recommend 0.502.00 as the
In summary, there is considerable support in
acceptability limits for duplicate QC ratios of
these data for temporary acceptance limits of 0.25
VOCs in groundwater. When these limits were
4.00 with the expectation of tightening in the fu-
applied to the most recent results for 67 samples
ture as procedures are improved. For the most
from 17 project locations, only 3% of the ratios
recent results for 19 samples from three project
were outside of this range.
locations, 13% of the QC/QA ratios were outside
The QC/QA ratio characteristics are summa-
0.254.00. This is similar to the original results
rized for nine VOCs in Table B7. Eleven of 321
and indicates a need for further improvement.
ratios (3.4%) were excluded because they were
Secondary standards prepared similarly to our
outside 0.303.00. Outliers were spread among
earlier recommendation for TPH should be used
several VOCs and they occurred over a wide spec-
throughout projects as a means of minimizing bias
trum of concentrations. Geometric means for each
between laboratories. We cannot isolate the con-
of the nine VOCs were close to the expected value
tributions of error sources such as sample hetero-
of 1.00, and the 95% confidence limits included
geneity and biodegradation in these results, but
1.00 in every case. The 99% tolerance bands are in
there clearly is a potential for improvement in
reasonable agreement with each other except for
this aspect of the program.
vinyl chloride, which offered insufficient data for
close tolerances (too few degrees of freedom).
Only three of the 310 ratios remaining after the
VOCs in groundwater
For the 157 sets of VOCs in groundwater, we
initial 11 exclusions were outside 99% tolerances.
discovered that four sets (V2-63 through V2-66)
As with metals in soils, let us examine the ef-
were repeats of V2-56 through V2-59. Thus, there
fect of using 0.402.50 as the acceptability crite-
were actually only 153 sets of results. For carbon
rion for QC/QA ratios. We find seven additional
tetrachloride, chlorobenzene, 1,2-dichloroethane,
exclusions beyond the 11 originally rejected val-
1,1-dichloroethene, and methyl ethyl ketone, there
ues for 0.303.00 limits. Of these 18 unacceptable
were fewer than five pairs of values where each
ratios (5.6%), 11 are for ratios below 0.40 and seven
was above the reporting limit. The number of pairs
are for ratios above 2.50. Further insight can be
above reporting limits for the other nine analytes
obtained from Table B8, which shows the distri-
ranged from nine to 63. Twenty-three of the QC
bution of ratios outside 0.402.50 with respect to
laboratory samples included concentrations for
the number of analytes above concentration re-
field duplicates, but results for most analytes were
porting limits for both QC and QA per sample.
below reporting limits.
Three samples have two outlier ratios each and
Only 1,2-dichloroethene and trichloroethene,
two of these three have no other analytes present
with 11 and 14 pairs of duplicate QC results, re-
above reporting limits. The other 12 rejected ra-
spectively, provided enough evidence to judge the
tios occur one per sample even though each of
limits to be expected for VOCs in groundwater.
these samples has acceptable ratios for one or
17
Previous Page
