Quality Control Report
The main purpose of this project was to disseminate valuable
data. The amount of data interpretation provided on the website is limited,
in order to allow users to customize the data to their own needs. However,
accurate data interpretation cannot be conducted without a thorough understanding
of the analytical quality of the dataset. The following section of this report,
also included on the website, provides a thorough review of quality assurance/quality
control (QA/QC) issues associated with each of the three studies incorporated
on the website.
It is important to note that, although each of the datasets
is of high quality, the original research programs were not required to conform
to Contract Lab Procedure (CLP)-type data standards, nor were the results
subject to an independent data quality review. Each project conformed
to the data quality requirements expected of the project when the research
was conducted. The Paleomagnetics and Environmental Chemistry Laboratory has
15 years of experience analyzing contaminated sediments, and routinely participates
in projects where rigid data quality standards are required. However, any
large dataset is likely to exhibit a small percentage of ambiguous data that
must be assessed individually by the analyst. Before any interpretations
are made with data on the website, users should carefully review the General
Laboratory Data Quality Objectives (DQOs) that were applied to these studies
(Appendix V), and have a thorough understanding of the Quality Assurance/Quality
Control (QA/QC) issues discussed below, and on the website.
The following QA/QC review is divided into two parts:
1) general data quality protocols and 2) a discussion of the quality of the
three individual datasets presented on the website:
A. Sediment quality in Rhode Island coastal
lagoons, 1999-2003
- Total digestion method for trace metal concentrations
in sediment: copper, lead, mercury
B. Sediment quality in Narragansett
Bay (including the Blackstone, Taunton,
Lee, Cole, and Kickamuit Rivers which
discharge into the Bay), 1997-1998
- Total digestion method for trace metal concentrations in
sediment: copper, lead, mercury
C. Changes in Narragansett Bay sediment quality
over time
A brief summary of QA/QC results
is provided in Table 1 at the end of the
report. This report provides a thorough review. A detailed discussion follows.
General Data Quality Protocols
|
|
All sediment data analyzed at the Paleomagnetics and Environmental Chemistry
Laboratory are required to adhere to a pre-defined set of QA/QC protocols.
All outlying data points are assessed individually before they can be included.
General methods used to assess data quality in this project are discussed
below.
Comparability:
All samples are collected, prepared, and analyzed
using established procedures that the laboratory has used over the past 15
years. These procedures have been thoroughly tested and used in several successful
intra- and interlaboratory comparisons. In addition, each procedure
is based on published and acceptable techniques.
Precision:
Duplicate sampling monitors the precision of the
analytical techniques. A duplicate sample is analyzed approximately
every 20 field samples, and the relative percent difference (RPD) between
the duplicates is used as an indicator of the reproducibility of the extraction
and analytical methods. Duplicates with RPDs of less than 30% represent acceptable
precision. Most duplicates are well below this percentage.
Accuracy:
The accuracy of analytical methods is evaluated
by analyzing Standard Reference Materials (SRMs) and Reagent Blanks. SRMs
are sediment samples obtained from the National Bureau of Standards that
have been certified for concentrations of 16 trace elements with a variety
of analytical techniques. In this project, at least one SRM is processed
with each batch of 20 field samples, and the resulting concentration of trace
metals is compared to the certified value for each analyte. If the
SRM concentrations are within 75-125% of the certified concentration, the
extraction and analytical techniques are considered to be accurate, and the
data are acceptable.
Reagent Blanks are samples that are composed only of the acids used in the
extraction process, and are analyzed to insure that there is no trace metal
contamination caused by laboratory techniques. Blanks are analyzed
with each batch of 20 field samples. Trace metal concentrations in all Blanks
must be below the laboratory reporting limit (similar to the instrument detection
limit) in order to conclude that no trace metal contamination has occurred
during the extraction process.
Both SRMs and Reagent Blanks are only used for trace metal analysis, and
are not applicable for grain size or organic carbon procedures.
Assessment of Outliers
Procedural Duplicates, SRMs, and Reagent
Blanks are considered representative of the batch of samples with which they
were analyzed. If all of these QA/QC samples are within the required
limits for each batch of samples, it is assumed that the data are accurate,
and no further assessment is required before reporting the data. If
one or more of the QA/QC results are outside of the required range, each
outlying data point is examined individually. There are a variety of reasons
why a data point might be outside of the QA/QC requirements, and it is up
to the analyst to determine the cause of the anomaly and the appropriate
course of action. For this report, all reported data are within the required
QA/QC ranges unless noted below.
It should be noted
that this report focuses primarily on data quality
protocols for trace metal analysis. We have
previously determined that our grain size and
organic carbon procedures consistently produce reproducible
results, and therefore we do not subject these
types of data to the same data quality review.
Similarity of Sampling Locations
For samples being compared between years,
verification of the location was important. Geographic coordinates
generated by LORAN-C and detailed field notes were recorded at each sampling
location during the 1988 survey. In order to return to the same location
in 1998, both GPS and the field notes were utilized. It is assumed
that locations do not vary by more than 100 meters.
Similarity of Extraction Techniques
It was necessary to insure
that the same sediment digestion method was used for each dataset.
Samples from 1998 study were processed using a "total" digestion method,
which is designed to fully liberate all metals from the supporting sediment
matrix. This method is known to produce the most accurate representation
of the concentration of trace metals in sediments. However, samples
from the 1988 study were processed using a "partial" digestion method, which
was a common extraction technique at that time. The partial digestion
method only releases metals adsorbed to sediment particles and organic material,
and can therefore result in trace metal concentrations that do not fully
represent the amount of trace metal contamination in the sample. In
order to overcome this problem, the samples taken in 1998 were re-digested
using the exact partial digestion procedure used for the 1988 samples.
For this project, only copper and lead concentrations that were analyzed
using the same extraction technique are compared.
Variation in Instrumentation
Samples collected in 1988 were analyzed
on ICP/AES (Inductively Coupled Plasma/Atomic Emission Spectrometer) and
Flame AA (Flame Atomic Absorption Spectrometer). Ten years later, samples
from the 1998 dataset were analyzed with the same ICP/AES (copper), but also
with Graphite Furnace AA (lead). In order to address any variability
in analyte concentration caused by instrumentation differences, the SRM used
during the 1988 study was reanalyzed with the samples collected in 1998.
Additional SRMs were also included with the 1998 samples. SRM recoveries
were comparable between the two datasets, indicating that variations in instrumentation
are not responsible for significant variability in analyte concentrations
between the 1988 and 1998 data.
It should be noted that this report focuses primarily on data quality protocols
for trace metal analysis. We have previously determined that our grain
size and organic carbon procedures consistently produce reproducible results.
Duplicates are not generally part of the standard operating procedure unless
there is sufficient sample remaining after all other analyses are completed.
See Appendix III for the Data Quality Objectives pertaining to these two
types of analyses.
Assessment of
Individual Datasets
|
|
A. Sediment quality
in Rhode Island coastal lagoons, 1999-present
(Total
Digestion Dataset)
There no QA/QC concerns
associated with this dataset. Samples were analyzed in two batches, with
the appropriate QA/QC samples (blanks, Standard Reference Materials, and
Procedural Duplicates) accompanying each batch. Duplicates were conducted
for grain size. Organic carbon was measured by loss on ignition (Dean,
1974) and verified against CHN. QA/QC results are excellent, suggesting
that the dataset accurately represents the level of trace metals, organic
carbon, and grain size in sediments.
B. Sediment quality in Narragansett
Bay, 1998 (Total Digestion Dataset)
This is a robust dataset,
with no major QA/QC concerns. Samples were analyzed in batches, with several
QA/QC samples (blanks, Standard Reference Materials, and Procedural Duplicates)
accompanying each batch. Overall, QA/QC results are excellent, suggesting
that the dataset accurately represents the level of trace metals in sediments.
As with any large dataset, a small number of QA/QC samples fall outside the
data quality objectives. These anomalies are discussed below, but did not
necessitate rejecting any samples. All analyzed data is included in this
project.
Copper
SRMs: Ten
Standard Reference Materials were analyzed for
copper. All but one were within the target
recovery range. The one outlying sample had
an extremely low recovery. However, two other
SRM samples analyzed in the same batch were characterized
by recoveries within the target range. The
low recovery for SRM 1646 was likely caused by an autosampler
pipetting error when the sample was analyzed, and is not considered
representative of other samples within the batch.
Lead
SRMs: Ten
Standard Reference Materials were analyzed for
lead. Two are above the target range by 2%
and 7%, respectively. Both problematic SRMs
were analyzed within the same batch, suggesting that
samples included in this batch could exhibit slightly
higher concentrations than is accurate. However,
an additional SRM in this batch was within the target
range, and the blanks do not suggest any evidence of contamination.
In addition, duplicates of three samples in the batch
in question were analyzed in other batches, and the correspondence
betweenduplicates is good. Therefore, the slightly high
recoveries for two SRM samples does not seem to indicate
a significant problem with the analysis, and no data were rejected
based on these two SRM recoveries.
Duplicates:
Eight samples were analyzed
twice for lead. One set of duplicates
exhibited a 43% difference between the samples,
which is outside of the target range. This
sample was analyzed in duplicate for several other trace
metals, and the percent difference between the duplicates
was outside the target range for approximately half
of the trace metals analyzed for this sample. This suggests
that sediment collected from this site is heterogeneous.
No data were rejected based on this one outlying duplicate.
Mercury
Blanks: Five
blanks were analyzed for mercury. One was higher
than the target concentration. The SRMs analyzed
with the batch in question do not show any evidence
of contamination. No data were rejected
based on this one outlying blank.
Duplicates: Eight procedural duplicates
were analyzed for mercury. One sample does
not conform to the data quality objectives. All other
QA/QC samples, including another duplicate, that were analyzed
with this group of samples are within target ranges,
so this outlying duplicate is not considered of concern.
C. Changes in Narragansett
Bay sediment quality over time
For this project, there is no mercury data. Changes in copper and lead
concentrations over time are documented by comparing 20 surface samples collected
in 1988 to samples collected at the same locations in 1998.
1998 partial
digestion dataset (surface samples):
This dataset
consists of 20 surface samples collected in 1998
and extracted using the same partial digestion technique
used in 1988 Samples were analyzed in batches,
with several QA/QC samples (blanks, Standard Reference
Materials, and Procedural Duplicates) accompanying
each batch. There were no major QA/QC concerns.
Any samples outside of the target range are discussed below.
Lead
SRMs:
Five SRMs were analyzed for lead, and two are
outside of the target range (61% and 74%). In general,
the partial digestion technique yields low recoveries
for lead, so it was expected that some SRMs would be
lower than the target range. The partial digestion technique
was used for these samples in order to insure that they
would be directly comparable to the 1988 dataset, and the
recoveries are similar between the two datasets. Therefore
no data were eliminated.
Duplicates: Three
procedural duplicates were analyzed for
lead. One pair exhibited a 65% difference between
samples, which is outside of the target range
of 30% difference.However, this sample is characterized
by a high percentage of sand (67%), which frequently
produces large disparities in analyte concentration.
An additional duplicate analyzed with this batch
is within the target range, so no data were eliminated due
to the problematic duplicate.
1988 partial digestion
dataset (surface samples and cores):
This dataset
consists of surface samples taken as part of the Narragansett Bay Project.
The dataset is divided into two groups: 1) surface samples from stations
7,12,15,17,18,19,20 collected in 1985; and 2) the remaining surface samples
collected in 1988. The QA/QC protocol was similar for this dataset
as for the 1998 dataset. However, the following differences should
be noted: A) For Group 1 above, Blanks and SRMs were analyzed, but duplicates
were not included due to a shortage of sample material; B) For Group
2, Blanks, Standard Reference Materials, and Duplicates were analyzed with
each batch of samples. However, SRM concentrations were not reported
for individual samples, but rather as an average of several SRM samples for
each analyte. Because only the average value was reported, it is possible
that a small number of the SRMs were outside of the target range.
The 1985-1988 dataset is considered to be comparable to the 1998 dataset,
and is very useful for documenting changes in trace metal concentrations
over time. However, ambiguities in the 1985-1988 SRM recoveries, particularly
for lead, suggest that small changes over time should not be considered significant
for the purposes of this project. Further studies will quantify the
specific samples associated with anomalous SRM recoveries, and the effect
on data interpretation will be addressed in more detail. Other individual
QA/QC concerns are documented below.
Copper
SRMs:
Group
1 samples:
Five SRMs were analyzed for copper,
with one (51% recovery) falling outside of the
target range. The original analyst appeared to have regarded
this SRM as an anomaly, possibly the result of laboratory
error, and did not eliminate any data in this batch.
Due to the careful attention paid to QA/QC during the
1985-88 project, this anomalous SRM is not considered to
be of concern for the current project.
Group 2 samples:
As mentioned above, SRM
recoveries were reported as the average of several
samples. This average recovery for copper is
83%, which is within the target range. However
it should be kept in mind that because this recovery is
an average, it is possible that a small number of individual
SRMs could have been outside of the target range.
Lead
SRMs:
Group
1 samples: In general, SRM recoveries
for lead are usually low when using a partial
digestion technique. For this project, 5 SRMs
were analyzed, with the following recoveries: 50%,
66%, 61%, 57%, and 132%. The same SRM was analyzed
with the 1998 partial digestion dataset, and yielded
74% recovery, which is nearer the target range than the
1985-88 samples. Although no data was eliminated
due to these low recoveries, it should be kept in mind
that lead concentrations from the 1985-88 dataset are
likely to be somewhat lower than those from the 1998 dataset,
and therefore the difference between the two datasets
could be somewhat exaggerated.
Group 2 samples: As mentioned
above, SRM recoveries were reported as the
average of several samples. This average recovery
for lead is 85%, which is within the target range.
However it should be kept in mind that because this recovery
is an average, it is possible that a small number of individual
SRMs could have been outside of the target range.
|
