more other analytes. The three samples with two
porting limits for all analytes, actual surrogate
outliers clearly require close scrutiny for likely
recoveries ("OK" is unacceptable), and to desig-
rejection of the entire analysis, but the 12 samples
nate method of analysis for each analyte.
with only one outlier may well be acceptable af-
2. A convention for significant figures should
ter examination of the whole set of analyses. The
be adopted for all concentration reports. One pos-
pleasing result is that 90% of the samples had no
sibility would be to use two significant figures
unacceptable ratios according to the suggested
for all values below 100 (e.g., 0.72, 6.9, and 54)
limits. Further support for using limits of 0.40
and a maximum of three significant figures for
2.50 for QC/QA ratios is gleaned from the less
values of 100 or greater. This would alleviate the
than 3% of outliers found in the most recent re-
problem of comparing pairs such as 6 and 6.25
sults for 67 samples from 17 project locations. The
(are they different?).
much greater homogeneity of groundwater com-
3. Standard practice should be adopted for re-
pared to soil and the reduced amount of manipu-
porting limits where dilutions are required. The
lation during preparation and analysis must cer-
target reporting limits given in ER 1110-1-263
tainly account for the vastly superior VOC results
(USACE 1990) might be specified as the highest
for groundwater.
acceptable values, assuming that these are low
Some VOC reporting limits were absent on the
enough for the intended use of the data. Special
data sheets and this reduced the effectiveness of
exceptions could be made for a few samples with
comparisons. The "practical quantitation limits"
unusual matrix problems. At the very least, both
for groundwater tabulated in Table D-4 of ER 1110-
QC and QA should specify reporting limits on
1-263 (USACE 1990) are 5 g/L for 12 of the VOCs,
the same basis rather than one on the original
10 g/L for vinyl chloride, and 100 g/L for me-
sample and one on a diluted portion.
thyl ethyl ketone. About three-fourths of the
4. A concerted effort is needed to reduce un-
samples apparently contained no analytes at con-
certainty in the QA values if they are to serve a
centrations requiring dilution prior to GC/MS.
referee purpose. This effort might combine the
For those samples, the ER 1110-1-263 criteria were
use of blind duplicates of selected samples, some
usually equaled or bettered. In general, however,
spike recoveries, and the regular analysis of certi-
the reporting limits for QA laboratories were lower
fied samples that are as similar in matrix as pos-
by about a factor of 2 than corresponding values
sible to regular samples. Such an expanded inter-
from the QC laboratories. Thirty-five samples
nal QC effort by the QA laboratories would help
listed one or both reporting limits above the ER
justify the use of their results for determining ac-
1110-1-263 recommendations, often by orders of
ceptability of analyses by other laboratories. Only
magnitude. More often than not the high report-
those laboratories with demonstrated excellence
ing limits were from the QC laboratory while the
should serve the QA function.
QA value was acceptable. Twenty-seven of these
5. A policy for dealing with outlier ratios and
samples contained one or more VOCs at concen-
unacceptable reporting limits must be promul-
trations high enough to have required substantial
gated by USACE. Such a policy could likely be
dilution. It appears that the QC reporting limits
incorporated into an artificial intelligence com-
were derived from diluted samples while QA val-
puter program to greatly reduce the time and cost
ues were for undiluted samples. From a practical
required to administer the policy.
perspective, the question is whether high report-
6. Consideration should be given to using as
ing limits decrease the acceptability of concentra-
few contract laboratories as possible and to avoid
tion estimates when one or more VOCs are present
using a given laboratory for the QA function on
at very high concentrations. Greater concern about
some projects and the QC function on others.
reliability would attach to the eight samples where
high reporting limits cannot be readily explained.
Metals in soils
Ratios of archived results for metals in split
soil samples were lognormally distributed for
CONCLUSIONS AND
those metals with sufficient numbers above con-
RECOMMENDATIONS
centration reporting limits (As, Ba, Cr, Pb). For
duplicate QC ratios, only 3.4% were outside the
General recommendations
limits of 0.502.00 while 10.2% of the QC/QA ra-
1. To improve future evaluations, both QC and
tios were outside limits of 0.402.50. Only five
QA laboratories should be required to submit the
samples out of 124 (4.0%) had more than one metal
method of estimation and numerical values of re-
18
Previous Page
