Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
March 17, 2015 ENCLOSURE 1
1
CONSOLIDATED AGENCY COMMENTS ON ERM DOCUMENT
“Data Quality Objectives Framework OU-1 Phase 1B RI”
Dated February 20, 2015
NOTE: These comments on the draft Phase 1B DQO document reflect discussions and
clarifications reached during scoping meetings held in Salt Lake City on March 11-12, and
reflect the main changes that the Agencies feel are necessary to ensure that the revised DQO
document is in accord with the discussions at the meetings. EPA reserves the right to provide
additional comments on the revised draft of the DQO document.
Major Comments
1. The DQO document uses an inconsistent and confusing nomenclature, often misusing the
term “COPC.” For example, stating that samples will be analyzed for COPCs is incorrect,
because contaminants of potential concern have not yet been selected. The DQO document
should be re-written to clearly state that all samples collected in Phase 1B will be analyzed
for the full list of chemicals provided in Worksheet #15 of the Phase 1A SAP. The list of
COPCs for the inner PRIs will be determined based on the results obtained from the samples
collected during the Phase 1B program.
2. As written, the DQO document implies that data sets of size 10-15 will provide UCL values
that will be adequate. This is not correct, and the text must be revised to clarify that UCL
values calculated from the data sets of this size MAY OR MAY NOT be adequate for well-
informed risk management decision making. Accordingly, the goals of the DQO presented
in Table 11-2 should be revised as follows:
Obtain sufficient data to allow for:
reliable identification of COPCs (human and ecological) in each PRI
calculation of 95% UCL values to support an initial assessment of human and
ecological risks
initial characterization of the nature and extent of site-related contamination
Do not use a phrase such as “reliable” or “acceptable” UCLs. All 95% UCLs (if calculated
correctly) are reliable and acceptable, in that they identify a value that has a 95% probability
of exceeding the true mean. Whether any specific UCL value is adequate for well-informed
risk-based decision-making can only be determined by performing the initial risk calculations
and identifying cases in which the risk conclusion based on the UCL is substantially different
from the conclusion based on the sample mean. In the absence of this process, the only way
to ensure that UCLs are adequate for decision making is to substantially increase the number
of samples collected.
March 17, 2015 ENCLOSURE 1
2
Accordingly, and in accord with discussions at the scoping meetings, the sample size for each
PRI should be revised to be a minimum of 14. The statistical justification for this sample
size should be revised to reflect the original COPC-based statistical assessment used in the
Phase 1A SAP, rather than an assessment of cost effectiveness (“diminishing return” in the
margin of error (MOE) between the UCL and the sample mean).
While not required, the DQO may also include an evaluation of the expected MOE between
the UCL and the sample mean for data sets of size 14, as a function of the expected range of
CVs. If presented, this should be based on an assumption that most data sets will be skewed
rather than normal. This is necessary because the historic data clearly demonstrate that
nearly all data sets from the inner PRIs are skewed, and many are well-approximated by a
lognormal distribution. A simple calculation of the expected MOE is not sufficient, because
this ignores the fact that both the sample mean and the sample standard deviation are random
variables, and hence the MOE is also a random variable. For example, if the true distribution
is assumed to be lognormal with a mean of 100 and a standard deviation of 250 (CV = 2.5),
the expected MOE is 2.94. However, if sampling variability is accounted for, the actual
distribution of MOE values is as shown in Figure 1. As seen, actual MOE values may be
much higher than the expected value. The rate of false positive decision errors is also quite
high for any COPC with a true mean greater than 0.1 times the RBC, as shown below:
Mean/RBC False Positive Rate
0.1 16%
0.2 37%
0.3 54%
0.4 66%
0.5 75%
0.6 82%
These calculations demonstrate and emphasize why data sets of size 14 may or may not be
adequate to support reliable risk management decision making, depending on the actual ratio
of the mean to the RBC and of the actual standard deviation.
3. The revised DQO document should include an independent 7-step DQO for a background
study of selected COPCs in soil. The goal of the study should be to identify one or more
upland and wetland habitat locations that are suitable for characterization of abiotic and
biotic concentration values of naturally-occurring or ubiquitous anthropogenic chemicals that
would be expected at the site in the absence of site-related contamination. The DQOs should
include a clear presentation of the null hypothesis (site > bkg), the statistical approach that
will be used to test the null hypothesis, and a demonstration that the sample size selected will
be sufficient to minimize the risk of decision errors. EPA recommends that the guidance
provided in EPA (2002) be followed. EPA expects that it should be possible to complete the
March 17, 2015 ENCLOSURE 1
3
text of DQO Steps 1-6 based on the discussions at the meeting. While Step 7 cannot present
final decisions about sampling locations, the text of Step 7 should discuss how suitable
candidate locations will identified and evaluated.
Also note that a background study of water (surface water and groundwater) is also needed.
EPA has agreed to provide initial thoughts on the DQOs and design of such a study during
one or more calls between the EPA and ERM toxicologists and during upcoming risk
assessor’s conference calls. A background water study should proceed according to a
timeline that informs Phase 2 collection of data in 2016 but implementation schedules are
still being contemplated.
Minor Comments
1. Page 3, middle paragraph. Use of the maximum value or the 95%UCL (whichever is lower)
as the EPC is no longer Agency policy. Selection of EPC values must follow current EPA
recommendations (see Section 1.10.2 of ProUCL Version 5.0.00 Technical Guide).
2. Page 3, last full paragraph. The following sentence is incorrect: “The existing set of
historical (pre-CERCLA) Inner PRI data have been judged by USEPA as not adequate for
use in risk assessments, including selection of COPCs and calculating 95UCLs.” While
EPA has determined that the historic data (alone) cannot be used for COPC selection or
UCL calculations, EPA has stated that, if historic data are found to be sufficiently
statistically similar to data collected in Phase 1B, it may be appropriate to combine the data
to increase sample size and decrease uncertainty. If so, then the historic data may be
included in the data set used for COPC selection and risk assessment.
3. Page 6, Section 11.3.2. When discussing the use of the LRMS method, include a brief
description (presumably detailed in the SOP) that summarizes when the method is expected
to be appropriate, and the triggers for re-analysis by HRMS (if needed). As per discussions
during the scoping meeting, criteria for reanalysis should be based on reliable quantification
of risk drivers, not high concentration low toxicity analytes. Refer to EPA’s letter of April
11, 2014, for EPA’s recommendations.
4. Page 8, Section 11.6. The use of the word “precision” to characterize the relation between a
sample mean and the 95% UCL of the sample mean is not correct. Precision is a measure of
the reproducibility between repeat analyses, while the relation between UCL and sample
mean is a measure of uncertainty. The document should be revised to use the term
“uncertainty” rather than “precision” to describe this relation.
March 17, 2015 ENCLOSURE 1
4
SUPPLEMENTAL COMMENTS ON THE PHASE 1A/B SAP AND ASSOCIATED SOPS
Although the following comments are not specifically relevant to the draft DQO submitted by
ERM, these comments are relevant to the Phase 1A/B SAP and associated SOPs that will be
required to achieve the DQOs.
1. When sampling sediment from water-covered locations in the lagoons, every effort must be
made to ensure that samples will be obtained from the target locations. Moving a sample to a
location that is accessible from the shoreline requires EPA approval.
2. The SAP for this project must establish a schedule that will ensure that all samples specified
in the DQOs for both the inner PRI study and the background study for soil and water (if
appropriate) will be collected, analyzed, and validated in time to allow thoughtful assessment
of the data in order to inform the requirements on any Phase 2 sampling or studies that may
be required in 2016.
CITATIONS
EPA. 2002. Guidance for Comparing Background and Chemical Concentrations in Soil
for CERCLA Sites. U.S. Environmental Protection Agency, Office of Emergency and Remedial
Response. 540-R-01-003, OSWER 9285.7-41. September 2002.
ProUCL Version 5.0.00 Technical Guide
http://www.epa.gov/osp/hstl/tsc/ProUCL_v5.0_tech.pdf
March 17, 2015 ENCLOSURE 1
5
FIGURE 1
DISTRIBUTION OF MOE VALUES FOR DATA SETS OF SIZE 14
DRAWN FROM A LOGNORMAL DISTRIBUTION WITH
MEAN 100 and STANDARD DEVIATION 250