QC/QA ratio per sample outside of the 0.402.50
We recommend that
limits.
a) As with VOCs in soils, temporary QC/QA
We recommend that
acceptance limits of 0.254.00 be adopted while
a) Acceptance limits for duplicate QC ratios be
procedures are being brought under better con-
set at 0.502.00.
trol, especially those at MRD. It is especially im-
b) Acceptance limits for QC/QA ratios be set
portant to ensure that both the QC and the QA
at 0.402.50. Application of these limits to recent
laboratories use the same extraction and analysis
results confirms that these are workable compari-
protocols depending on the requirements of the
son criteria for both QC1/QC2 and QC/QA.
applicable state method (if present).
c) Future results for multianalyte procedures
b) Secondary standards sealed in ampoules
should be carefully analyzed with regard to the
should be prepared and used to identify sources
number of rejects/sample. From such analysis, a
of systematic error so they can be reduced to a
policy for dealing with such analyses should be
minimum.
developed.
c) A nested experimental design should be used
to evaluate the magnitude of errors in sampling,
sample homogenization and splitting, and stor-
VOCs in soils
Archived results for VOCs in split soil samples
age. This information would be used to focus at-
produced QC/QA ratios that were distributed log-
tention on any large random error sources.
normally as expected. However, over 40% of the
QC/QA ratios were outside the range 0.254.00
Explosives in soils
and 24% were outside 0.1010.0. The magnitude
Ratios of MRD archived results for explo-
of this scatter is so large that it is impossible to
sives in split soil samples analyzed by QC and
recommend effective limits of acceptability. In-
QA laboratories were distributed lognormally,
stead, we believe that steps are urgently needed
with 89% of the 65 QC/QA ratios between 0.25
to improve data quality.
4.00 and 72% between 0.502.00. Because many
We recommend that
concentrations were close to reporting limits,
a) Temporary acceptance limits of 0.254.00 be
there were 84 cases where either the QC or the
set, despite the possibility of up to 40% unaccept-
QA laboratory reported "less than" while the
able data. Analysis of recent results using these
other provided a real number. For these 84 pairs,
limits indicated some improvement, but more is
63 (75%) were considered to be in acceptable
needed.
agreement.
b) Secondary standards sealed in ampoules (see
The NPD submitted a series of analyses from
Hewitt 1994, Hewitt and Grant 1995) should be
the Umatilla Washout Lagoon where concentra-
used for performance evaluation during labora-
tions of RDX and TNT were very high. There was
tory qualification and as periodic QC samples.
a large bias (nearly a factor of 2) between the QC
c) The change with the greatest potential for
and QA laboratories, possibly due to either cali-
data quality improvement would be to start a pro-
bration differences or incomplete extraction by the
gram of improved techniques for sample collec-
QC laboratory.
tion, storage, and laboratory subsampling in
We recommend that
preparation for analysis. Specific recommenda-
a) Temporary acceptance limits for QC/QA ra-
tions can be found in Hewitt et al. (1995). An
tios be set at 0.254.00. Application of these limits
instructional videotape might prove to be useful.
to recent results yielded 13% of the ratios outside
this range. We believe this percentage can be sig-
nificantly reduced.
TPH in soils
Ratios of archived results for TPH in split soil
b) Secondary standards containing known con-
samples analyzed by QC and QA laboratories
centrations of explosives in dry soils and sealed
were provided by the North Pacific Division
in ampoules should be used by participating labo-
(NPD) and also by the Missouri River Division
ratories on a regular basis in order to reduce
(MRD). Nearly half of the QC/QA ratios for the
interlaboratory bias.
MRD results were unacceptably wide, so only the
c) Sample collection, splitting, and storage pro-
NPD data was statistically analyzed. Although
cedures should be reviewed with the objective of
86% of these ratios were between 0.254.00, the
reducing variations, especially biodegradation
standard deviation of the logarithms was larger
losses of nitroaromatics.
than expected for a lognormal distribution.
19
Previous Page
