EPA/600/R-16/113 | June 2016 | www.epa.gov/ord
United States
Environmental Protection
Agency
EPA Gold King Mine Analysis
of Fate and Transport in the
Animas and San Juan Rivers
Response to Peer Review Comments
Office of Research and Development
National Exposure Research Laboratory
-------
Response to Peer Review Comments of EPA's
Gold King Mine Analysis of Fate and Transport in the
Animas and San Juan Rivers
U.S. Environmental Protection Agency
Office of Research and Development
Washington, D.C.
July?, 2016
-------
Disclaimer
This document has been reviewed in accordance with U.S. Environmental Protection Agency
policy and approved for publication. Mention of trade names or commercial products does not
constitute endorsement of recommendation for use.
-------
Table of Contents
OF
I. Introduction 4
II. CHARGE TO REVIEWERS 6
III. GENERAL IMPRESSIONS 8
IV. RESPONSE TO CHARGE QUESTIONS 16
Part 1: Overall Project and Analysis 16
Question 2 21
Questions 26
Question 4 28
Part 2: Fate and Transport 31
Question 5 31
Question 6 39
Question? 44
Questions 49
Part 3: Geochemistry 50
Question 9 50
Question 10 51
Question 11 52
Part 4: Water Quality Analysis Simulation (WASP) Modeling 53
Question 12 53
Question 13 63
Question 14 68
Part 5: Groundwater Modeling 70
Question 15 70
Question 16 81
Question 17 84
Question 18 86
Part 6: EnviroAtlas Modeling 88
Question 19 88
Question 20 90
Part 7: Bioaccumulation 91
Question 21 91
-------
I. INTRODUCTION
The Office of Research and Development, National Exposure Research Laboratory has been
conducting an analysis of the release of acid mine drainage from the Gold King Mine on August
5, 2015 and its transport and fate within the Animas and San Juan Rivers. This project's
objectives are to provide analysis of water quality following the release of acid mine drainage in
the Animas and San Juan Rivers in a timely manner in order to 1) generate a comprehensive
picture of the plume at the river system level, 2) help inform future monitoring efforts and 3) to
predict potential secondary effects that could occur from materials that may remain stored within
the system. The project focuses on assessing metals contamination in the rivers following the
release of metals from the mine and during the movement of the plume and in the first several
months following the release. A quality assurance project plan was developed for the work in
this project.
A mid-project peer review was managed by Versar, Inc., an independent contractor, under
contract No. EP-C-12-045 Task Order 80. Versar was tasked by EPA to coordinate an external
peer review of EPA's project entitled Gold King Mine (GKM) Analysis of Fate and Transport in
the Animas and San Juan River. The purpose of the three-day peer meeting, held at the EPA's
Office of Research and Development (ORD) Laboratory in Athens, Georgia on February 23-26,
2016, was for five expert reviewers to evaluate the scientific integrity of EPA's analysis and
characterization of the fate, transport, and potential impacts of acid mine drainage (AMD)
release in the Animas and San Juan Rivers. The reviewers met with EPA scientists who
presented their analysis and findings to the reviewers via Power Point presentations that had
been provided prior to the meeting. The peer review process provided a documented,
independent, and critical review of the draft analysis, and its purpose was to identify any
problems, errors, or necessary improvements to the analysis prior to being published or
otherwise released as a final assessment. Project findings included in a final report will also be
independently peer reviewed.
Versar was charged with assembling the peer reviewers and coordinating the peer review.
Versar evaluated the qualifications of peer review candidates, conducted a thorough conflict of
interest (COI) screening process, independently selected the five peer reviewers, distributed
review materials, managed the written peer review period, organized and hosted the peer review
meeting. Versar identified candidate reviewers with expertise in the following areas: (1)
geochemistry, (2) fate and transport (water/sediment), (3) water quality analysis simulation
(WASP) modeling, (4) groundwater modeling, (5) geospatial analysis (EnviroAtlas modeling),
and (6) bioaccumulation. Versar's in-depth and multi-staged evaluation of qualifications was
based on each candidate's biosketch, curriculum vitae (CV), and publications.
Peer reviewers included:
Brian S. Caruso, Ph.D., P.E.. U.S. Geological Survey, Denver, CO
Charles R. Fitts, Ph.D., Fitts Geosolutions, LLC. Scarborough, ME
HenkM. Haitjema, Ph.D., Haitjema Consulting, Inc., Bloomington, IN
D. Kirk Nordstrom, Ph.D., U.S. Geological Survey, Boulder, CO
William A. Stubblefield, Ph.D., Chair, Oregon State University, Corvallis, OR
-------
Versar developed a final peer review report that summarized the peer review comments
provided during the meeting and presents the reviewers' individual written comments in
response to a series of charge questions pertaining to hydrology, geochemistry, fate and
transport, and potential impacts from the Gold King Mine release (Versar, Mar 9 2016).
This EPA document contains the EPA response to the independent peer reviewer's comments
provided in Versar's summary report. In some cases, the reviewer's offered comments or
opinions that were outside the scope of the charge questions and this scientific project. This
document responds only to comments directed to the technical aspects related to the ORD
analysis.
Comments were prepared by the Office of Research and Development Gold King Mine Project
Team:
Kate Sullivan, Ph.D
Christopher Knightes, Ph.D.
John Washington, Ph.D.
Mike Cyterski, Ph.D.
Steven Kraemer, Ph.D.
M. Craig Barber, Ph.D.
Anne Neale
Lourdes Prieto
-------
II. CHARGE TO REVIEWERS
Part 1. Overall Project and Analysis
Question 1. Given the data that were available to the researchers at the time, were assumptions
about data inclusion, formatting, and use appropriate? How so?
Question 2. Was the overall integration process of the various analyses conducted in a way that
provided meaningful results and conclusions? Please explain.
Question 3. When looking at the full project, are there errors or gaps in the integration process
that could have affected the overall analyses and/or the conclusions? Please explain.
Question 4. Were the overall conclusions that were drawn from these analyses appropriate and
scientifically defensible based on the analysis? Why or why not?
Part 2. Fate and Transport
Question 5. Does the research appropriately characterize the metals concentrations and load
produced at the Gold King Mine spill?
Question 6. The concentration of metals near the release site in the receiving waters had to be
estimated from samples collected after the much of the plume had passed. Were the estimates of
metals concentration at this location appropriately calculated through scientifically sound
methods using available data?
Question 7. Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the Animas River?
Question 8. Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the San Juan River after receiving
mine contaminated water from the Animas River?
Part 3. Geochemistry
Question 9. Were the geochemical principles to characterize transport and fate of acid mine
drainage appropriately applied and interpreted? Please explain.
Question 10. Were precipitation and mineral saturation analyses of the acid mine drainage
appropriately applied for interpreting metals fate in the river system? Please explain.
Question 11. Was the neutralization of acid mine drainage and subsequent fate of dissolved and
colloidal/particulate metals appropriately interpreted? Why or why not?
-------
Part 4. Water Quality Analysis Simulation (WASP) Modeling
Question 12. Did the WASP modeling appropriately apply modeling parameters to estimate the
movement of plume water? Please explain.
Question 13. Did the application of assumptions and values in WASP modeling appropriately
address particle transport and deposition of the acid mine drainage constituents? Please explain.
Question 14. Did the WASP modeling appropriately investigate the remobilization of metals
during increased flow? Why or why not?
Part 5. Groundwater Modeling
Question 15. Is the analysis as presented sufficient to evaluate the potential for impact of the
acid mine release from the GKM on pumping wells located in the floodplain aquifers
downstream of the spill?
Question 16. Were the assumptions informing the choice and construction of the groundwater
flow model appropriate for the intended use? Please explain.
Question 17. Were the assumptions informing the capture zone and particle tracking analysis
appropriate for the intended use? How so?
Question 18. Did the method for calibration of the local scale groundwater flow model
performance to the observed drawdown reported in the driller's log serve as an effective
method? Please explain.
Part 6. Atlas Modeling
Question 19. Are the sources of the data included in the maps valid, complete, and adequately
documented? Are there any points of confusion, gaps, or suggestions for improvement?
Question 20. Do all of the maps and charts communicate the analysis methods and results in
such a way as to be readily understood by stakeholders with interest in the impacts of the Gold
King Mine spill (e.g., First Nations; NGO's; news media; and State water, recreation, public
health, and wildlife managers)? Are there points of confusion, gaps or suggestions for
improvement?
Part 7. Bioaccumulation
Question 21. Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, and Zn in the Animas River trout fishery? What are the
strengths and weaknesses of using this approach?
-------
III. GENERAL IMPRESSIONS
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
1)1 commend EPA for gathering and analyzing all of this data in
an attempt to understand the contaminant plume movement in the
Animas and San Juan rivers from the Gold King Mine release. It is
always challenging to collect and analyze in a consistent way
existing data from a wide range of sources and with different
levels of QA/QC. EPA has done a relatively good job in a short
time frame at a first cut for this fate and transport analysis.
However, the accuracy of information presented is questionable
due to a number of reasons and assumptions, the clarity of
presentation needs improvement, and the soundness of
conclusions is also drawn into question based on these issues. One
of the main issues is that the goal of the research appears to be too
broad and not specific enough to determine if the information and
conclusions are adequate. In some cases the goals and objectives
are not entirely clear and appear to be somewhat different in
various places in the presentation where they are presented.
2) In general, I believe that EPA should perform this work and
prepare the research analysis so that it uses the best science
available and presents results as clearly as possible in preparation
for a number of issues, including potential lawsuits and Superfund
investigations, monitoring plan development, and to inform all
stakeholders of what occurred as best as possible. Although many
of the conclusions seem generally appropriate based on the
analyses performed, the quantification and accuracy of the
conclusions are weak due to a great deal of missing information
and lack of detailed uncertainty and sensitivity analysis.
1) The primary objective of EPA/ORD's Gold King
Mine study was to quantify the amount and
characteristics of AMD produced in the GKM release
and its transport and fate in the Animas and San Juan
Rivers during the plume and in the immediate period
following the initial event. The scope of the project is
specific and supported by data collected by EPA,
states, and tribes during and following the event.
Each analysis within the project supports the overall
objectives with specific objectives. We will further
clarify objectives and integration of analyses in the
final report.
2) A basic principle of EPA is to conduct the highest
quality science and to ensure this through peer
review. This mid-project peer review contributes to
meeting those goals. We will revise the organization
of materials and visualization of results to more
clearly communicate study objectives, methods, and
findings in the final report. We will also include
uncertainty and sensitivity analysis.
-------
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
3) Several examples of where the analysis and presentation should
be improved include:
better definition of goals and objectives to reflect critical
information needs
use of EPA national criteria or standards for metals for
drinking water and aquatic life as part of an initial
screening of risk to select potential contaminants of
concern for more detailed analysis and as indicators
(instead of primarily evaluating total metals)
better use of other existing data and information from
previous investigations to evaluate and help confirm
background (pre-release) levels for comparison
inclusion of additional data and information for better
reactive transport modeling, metals concentration and load
calibration, and validation for WASP
better evaluation and presentation of uncertainty and
sensitivity analysis of results should identify data gaps in
the analysis and for future modeling
3). Our revised materials and report will include a
clear expression of goals and objectives. In addition,
the final report will include more detailed treatment
of the topics identified by the reviewer, including:
evaluation of the plume from an exposure
perspective; analysis of pre-and post-GKM event
water and sediment concentrations, additional data
analysis for calibrating WASP (see questions 12 -
14). We will also increase discussion of uncertainty,
model sensitivity, and data gaps.
Charles Fitts
4) It is hard to summarize since there are so many facets of these
studies. The soundness of conclusions is discussed under question
4 below. There are many details that need attention, and many of
these are just a matter of editing, polishing and fleshing out with
more text and detail, which is to be expected in a more final draft.
I felt that the overview and empirical analysis sections were
generally logical and needed minor work. I have few comments on
the geochemical and bioaccumulation portions since I have less
background in those areas. The WASP presentation could use a
good deal of clarification about the analyses and more caveats
about the uncertainties involved and how the results may be used.
4) As part of the WASP calibration activities we
have significantly improved the ability of the model
to perform deposition/suspension analysis that
reasonably reflects the observed data and that squares
with published data. See responses to questions 12-
14 for more discussion of these points.
We have included more sophisticated 3D
groundwater modeling to augment the 2D modeling
that was reviewed by the panel. We have also
gathered additional data for a greater number of
-------
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
For example, the deposition/suspension analysis of WASP slides
25-27 seems to be quite uncertain and should be viewed
skeptically since the deposition/suspension input parameters do
not square with published ranges. In most cases, the WhAEM
modeling was sufficient to characterize whether a well likely pulls
in some river water, but the modeling approach was not
sophisticated enough to predict accurately what fraction of a
well's flow came from the river and what the plume breakthrough
curve might look like in well concentrations. More sophisticated
3D and localized models could be constructed to improve
predictions, but the Key Analysis Question (Groundwater slide 2)
may not require such detail for most of the wells.
wells in the vicinity of the modeled wells to enhance
calibration to river and well water levels. These
responses are described in greater detail in a
following section (Part 5 questions 15-18)
responding to specific feedback on groundwater
modeling.
HenkHaitjema
5) The overall goal of the research has been the topic of some mild
confusion by me (and the group at large). The agency stressed that
the current research does not constitute a formal risk assessment
nor was it designed as such. However, the precise purpose of the
research has not been articulated very clearly. I must assume that
in the end the research presented is to be used as the basis for
some form of risk assessment and, if needed, remedial action. As
such I have been evaluating the research presented with this
ultimate goal in mind.
6) Overall the work was well presented although the complexity of
some issues and the necessary brevity of the presentations resulted
in many interruptions of the presentations with questions or
requests for clarifications by the reviewers. While I understand
that the EPA researchers could only work with publically available
data, it was observed by several reviewers that some important
historic (background) data were missing, but might have been
acquired from public sources (e.g. the USGS).
5) The project objectives were to quantify the
volume and type of metals introduced into the
Animas River system from the Gold King Mine
release, the water quality characteristics during
transport, and the fate of the metals in the receiving
rivers. The final report will present results including
the magnitude and duration of potential exposure of
water users and assessment of possible future effects.
6) The EPA team has acquired publicly available
historic data from the USGS and EPA STORET
databases as well as data available from states and
tribes to enable an expanded comparison of metals
concentrations pre- and post-event in surface water
and sediment.
10
-------
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
7) Most conclusions seemed reasonable, taken into account the
limited data and the basic nature of these initial studies. However,
in several cases the data uncertainty could have been better
alleviated with some sensitivity analyses and by presenting
bracketing solutions showing both most favorable and most
unfavorable (worst case) scenarios.
8) Finally, I have the impression that communication between
different branches of EPA is less than optimal. On several
occasions the quality of the studies suffered as a result. For
example, the lack of coordination between the various sampling
efforts and the lack of information about the sampling and quality
assurance protocols cast some doubt on data integrity, hampers
data comparisons, and may have resulted in unnecessary data
gaps.
7). The final report will include increased treatment
and presentation of uncertainty and sensitivity for
empirical and modeling analyses.
8). The EPA/ORD team utilizes data developed by
EPA Regions 8, 6, and 9 that were collected and
processed following Quality Assurance Project Plans
and SOPs that document sampling protocols, testing,
and quality assurance. We also included data from
the USGS, the states of Colorado, New Mexico, and
Utah as well as the Southern Ute Indian Tribe and
the Navajo Nation who have similar QA/QC related
documentation.
Kirk Nordstrom
9) The Animas River Team (ART) of the EPA's Office of
Research and Development (ORD) involved with research on the
fate and transport of potential contaminants from the Gold King
Mine (GKM) spill presented, summarized, and interpreted a very
large set of diverse data collected by EPA and other technical
groups under adverse conditions. Although the data set was large,
many necessary parameters were missing and the quality was less
than optimal for the objectives of the ART because the accidental
release was unexpected and field and lab parameters were
collected while the EPA was in an emergency response mode with
little time for planning. Hence, the ART was working under a
serious handicap and with very tight time constraints. Considering
this overall situation, the presentations were impressive. They
have made every effort to be thorough in collecting information,
careful in most of their decisions on how to proceed with
insufficient data, and they have been clear on what information is
9). See response (8). The EPA/ORD team has
utilized available data to the fullest extent possible.
Although there are some data gaps, such as several
geochemistry-related parameters not routinely
measured in field sampling in that would have
enhanced geochemical analysis, the EPA/Ord team
believes that the cumulative data set compiled from
multiple agencies allows a robust analysis of pre-,
during, and post event metals contamination by the
Gold King Mine event in the 600 km of river
sufficient to meet the objectives of this project which
was to understand the source, transport and fate of
the GKM plume metals. The source and use of data
in analyses will be fully described in the final report.
11
-------
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
based on fact, what assumptions were used, what aspects were
largely speculative and require follow-up monitoring, and they
have reviewed and revised many of their conclusions to keep them
as sound as possible. They have been transparent about what they
have tried to do and completely open to good suggestions. We
had, in my opinion, excellent discussions about what can and
cannot be done with the available data.
10) That's not to say that there isn't room for improvement. To be
sure, some of their assumptions could use revision, some of the
methods that were used need modification, and in one instance
(bioaccumulation) the effort was highly questionable. Having an
independent review to evaluate the work at this point was a wise
choice. As long as the recommendations of the reviewers are
carefully considered, this mid-point evaluation should prove
extremely valuable in helping the ART to achieve its goal.
11) Some of the figures in the presentations were impossible to
read either in the hard copy or in the various PowerPoint
presentations. These should have been checked and improved.
12) A more logical and consistent sequence to the presentations
would have helped also. A more helpful logical and consistent
sequence means a clear statement of goals followed by an outline
of available data with a tabulation of the logic on how to obtain
said goals. Some of this was presented but it was a bit different for
each group and the methodology was not always clearly stated.
13) It is difficult to appropriately characterize the metals
concentrations and loads when a lot of the important field and lab
data were not collected. Immediate field reconnaissance was
10) The EPA/ORD team agrees that that this mid-
project review by independent experts was very
beneficial. Reviewers' comments helped us improve
our methods and presentation of information.
Individual subject experts agree that the peer review
team was deeply insightful in the scientific areas
covered in this project and we appreciate the time
and effort they gave us and the in-depth and lively
discussions of our work. The review comments have
strongly influenced the trajectory and details of all of
the analyses and subsequent presentation of data that
will be included in the final report. These will be
further highlighted in response to specific comments
that follow.
11). Figures and data presentation will be carefully
prepared in the final report for clarity and lack of
errors.
12). The final report will be organized to integrate
findings along topic areas. The peer review was
organized to review methods. We will include a table
of approaches and other means do display the
individual study elements and their integration.
13) When collecting water samples that were
analyzed in the laboratory for dissolved and total
metals, EPA field crews routinely measured water
temperature, pH, specific conductance and dissolved
oxygen.
12
-------
General Impressions
Reviewer Name
Reviewer Comment
challenging because of the unexpected accidental and sudden
release of mine pool water, the time delay in notifying authorities
of the accident, and the time delay in getting personnel and
equipment to the field. Of course, under rapid emergency
conditions it is difficult to collect enough of the right kind of data.
However, it is hard to understand why more field parameters were
not measured such as conductivity, pH, and temperature for all
samples, why sulfate and Fe(II/III) were not determined when
water samples were collected for analyses, and why no samples of
GKM effluent were collected during the release. These parameters
(pH, conductivity, and Fe(II/III) should always be measured for
acid mine water contamination. This is not a criticism of the ART
modeling efforts, obviously, but of the lack of guidelines for the
field personnel who collected the samples.
14) The EPA should have a handbook that recommends what
samples and field parameters need to be collected in an emergency
mine water spill. Furthermore, the handbook should emphasize the
importance of getting water samples of the source water (the Gold
King Mine effluent) as soon as possible and throughout the main
pulse of mine water release because its chemical composition
could, and probably did, change during the release. It is imperative
that the chemical composition of the pollutant source be properly
characterized because substantial changes in its composition can
occur and will affect downstream transport. If the source is not
well characterized then it becomes extremely difficult for the team
to characterize the changing conditions of the plume as it moves
hundreds of kilometers downstream. If the proper parameters had
been collected, the ART could have done far better at
characterizing the metals and the load, the rate of movement of the
plume, the partitioning of metals between dissolved and
EPA Response
14) The EPA/ORD team agrees that additional data
would be useful for sophisticated geochemical
analysis of an AMD plume such as occurred with
Gold King Mine. These would include (in generally
decreasing order of importance): i;) for field pHs
above 4, simple field alkalinity titration; ii) dissolved
Fe speciation, ferrous vs. ferric; iii) anions,
especially SO4=; iv) for pHs less than 5, laboratory
acidity, with the sample pre-oxidized with FhCh and
titrated over heat to accelerate reactions and
SEM/AVS extraction as a metric of potential toxicity
to aquatic organisms.
We note that a limited amount of this type of
geochemically-relevant data such as sulfate was
collected by the states of Colorado, New Mexico and
Utah as well as the Southern Ute Indian Tribe that
were used in the analysis.
13
-------
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
particulate forms, and the fate of the metals in the plume. What the
Team did manage to do with this partial data set is highly
commendable, appropriate, and the results were very reasonable.
More on this below.
15) A general rule of thumb is that anybody trying to model the
hydrogeochemical dynamics of a field site needs to see the field
site. By visiting the sites, the team can get a much better idea of
how appropriate their modeling and assumptions are for the goals
of the project. I was surprised that no one had been allowed to see
the area or had ever visited the area. A good field observer has a
natural feel for how to model a complex and transient event with
limited data. This disconnect between field and modeling effort
can lead to inappropriate analyses and conclusions.
15). The EPA/ORD team agrees that direct field
observation helps researchers gain insight when
conducting analysis. However, the research team
was able to use resources to help fill the gap.
Members of the EPA/ORD team have had extensive
field experience with AMD and hydrology and water
quality measurement as well as with use of the
models we deployed. We consulted with EPA
regions onsite and were informed by other
researchers and agency reports who conducted onsite
sampling We were able to use remote sensing
imagery to view the entire system. There were
abundant photographs, news videos and images that
provided visual impressions that were as good for
many purposes as direct observation.
William
Stubblefield
16) The presentations provided in the Gold King Mine Analysis of
Fate and Transport in the Animus and San Juan Rivers reflected
the high degree of effort and quality expended in their preparation.
However, the overall objectives of the effort, technical approaches
employed, and desired outcomes were not obvious. EPA NERL
scientists were clearly at a disadvantage not having been involved
in the design of the sampling plan, its implementation, and the
assessment of the overall quality of the data. Two analysis
objectives were stated in the overview presentation:
Characterize the release, transport and fate of the
approximately 3 million gallons of released AMD, with a
focus on a suite of metals
Identify the potential for water quality impacts, including
municipal wells, and implications for future monitoring
priorities.
16) The primary objective of ORD' s Gold King
Mine study was to quantify the amount and
characteristics of AMD produced in the GKM
release, and its transport and fate in the Animas and
San Juan Rivers during the plume and in the
immediate period following the initial event. We
considered the potential for groundwater routes of
exposure. We also attempted to provide some insight
on what might happen to metals sequestered in the
system in future runoff events. This analyses has
been expanded since the peer review and will be
included in the final report.
14
-------
General Impressions
Reviewer Name
Reviewer Comment
EPA Response
17) Clearly, a great deal of effort went into addressing the first
objective and EPA scientists did a reasonable job of achieving this
objective, given the limitations in data and the rapid nature of the
response. It is not as clear how the second objective was to be
addressed. Prior to the review it was explained that this was not an
"ecological risk assessment;" however, to be able to address the
second objective it is imperative that environmental exposures for
individual metals be adequately described in terms of their
magnitude and duration, as a minimum. Given the current state-of-
the-science it would also be helpful to have information regarding
those physicochemical parameters that can affect the toxicity of
individual metals to aquatic organisms (e.g., dissolved organic
carbon, pH, and hardness). It was also noted that there was a
reasonable set of sediment data analyses (300 samples) but no
detailed analysis of this data was presented. It was acknowledged
that there is a large amount of data available and that the
integration and interpretation of the data represents an onerous
task, especially given the rapid analysis time available.
18) In conclusion, it was somewhat difficult to discern what the
objectives of the integrated program were and whether they had
been achieved or not. There seemed to be a lack of cohesiveness in
the overall program objectives and the approaches taken to
achieve these objectives.
17) The final report will contain information on the
magnitude and duration of environmental exposures
during the movement of the Gold King Mine plume
through the affected rivers.
The final report will contain the available data on
physico-chemical parameters including hardness and
pH that were collected during plume travel. There are
no known organic carbon data collected during the
event.
The final report will also present data and analysis on
metals concentrations in streambed sediments and
the effect of the plume on them.
18) See responses (5) and (16). The final report will
be organized to integrate findings along topic areas.
We will include a table of approaches and other
means to display the individual study elements and
their integration.
15
-------
IV. RESPONSE TO CHARGE QUESTIONS
Part 1: Overall Project and Analysis
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
19) Some assumptions about data inclusion, formatting,
and use were appropriate, some were not, and some were
questionable. There appear to be many questions and
issues with regard to the analysis methods and
assumptions, many of which affect our evaluation of the
assumptions about data inclusion, formatting, and use.
Important questions and issues include:
a) The goals of the fate and transport analysis and
modeling are not clear, and in some cases appear
to be different in various parts of the presentation
materials.
b) It appears that the WASP TOXI model for
toxicants, including metals, was not used. This
module incorporates Kd values for partitioning
between dissolved and particulate forms, 1st order
decay, and diffusion coefficients, for some
reactive transport modeling. Also, why was the
WASP add-on, Metals Transformation and
Assessment (META4), not used for the fate and
transport modeling? This module was developed
by EPA and can handle reactive transport in
complex acid mine drainage-metals systems with
precipitation-dissolution reactions incorporating
pH and other important parameters.
19)
a) The project objectives were to quantify the
volume and type of metals introduced into the
Animas River system from the Gold King Mine
release, the water quality characteristics during
transport, and the fate of the metals in the receiving
rivers. Each of the analytical methodologies
reviewed individually had a more specific way of
framing that charge depending on the subelement
and nature of the method. The final report will
ensure clarity on project objectives and scope.
b) We used the WASP TOXI module for the fate
and transport simulations. This was not made
specifically clear in our presentations. We state this
explicitly now. We have updated our model
formulation following discussions with the
reviewers. We have incorporated diffusion of
dissolved species between water column and pore
water and a lumped parameter Kd for individual
metals for partitioning to and formation of
particulates. We also have used the empirical
estimates of total loads at selected locations to
estimate constant settling velocities. We reached out
to Dr. James Martin, one of the WASP architects, to
ask about META4. META4 was used on mine sites
in CO by Medine in the 1990s, which is based on
16
-------
Reviewer Name
Reviewer Comment
EPA Response
c) Important or indicator individual metals should
be analyzed and presented in more detail. These
should probably include at least Cd, Cu, Pb, and
Zn. Summary statistics of data should be
calculated at time periods along the length of
rivers and compared to EPA drinking water and
aquatic life hardness-based criteria to evaluate
potential contaminants of concern for fate and
transport analysis and initial screening for
potential risk.
d) Why did EPA not use Sondes for continuous
monitoring of parameters such as conductivity and
pH?
e) Why was pH and conductivity not measured in
many samples?
f) Why were different sampling and analysis
methods and detections limits used by different
EPA organizations and for different samples?
WASP4. To use META4, the analyst would need to
know META4 as well as WASP4. We are currently
using WASP7, which has had many upgrades and
updates since WASP4. META4 was never officially
released. The routines available in WASP7, such as
the new hydrology routines, which allows for the
kinematic wave modeling for streams/rivers or the
interface, are not be available in WASP4. An update
to META4 was done in 2012, though it has never
been fully implemented into WASP, and is not
currently available or usable. Dr. Martin
recommended that META4 not be used until it is
finally incorporated into the updated WASP, and
then only if there is good water chemistry data.
c) The final report will include analysis of
concentrations and mass of individual medals
including at a minimum Cd, Cu, Pb, and Zn at
various points along the rivers. These
concentrations will be evaluated against various
metals-related water quality criteria.
d) Continuous sonde data on pH and specific
conductance was collected by the state of New
Mexico and the Southern Ute Indian Tribe at several
locations in the lower Animas River and near
Farmington in the San Juan River. The EPA/ORD
team used these data in establishing plume
movement.
17
-------
Reviewer Name
Reviewer Comment
EPA Response
e) EPA crews measured specific conductivity and
pH when collecting samples. These data are
included in analyses.
f). EPA samples were collected and laboratory tested
according following Quality Assurance Project Plans
in place in each Region. Laboratory testing was
consolidated under one national laboratory several
days into the GKM release, which accomplished
standardization of reporting results.
Charles Fitts
20) There was some discussion about other possible
sources of data from academics and other organizations.
If there exist other data particularly at an earlier time near
the GKM or Cement Creek, it would be helpful to get that
data and include it in the analysis, since it would reduce
the uncertainty about the source concentrations and mass.
20) The EPA/ORD team has obtained pre-event data
from the USGS and EPA STORET data bases and
all available data pre-, during, and post the GKM
event from the states of Colorado, New Mexico, and
Utah as well as the Southern Ute Indian Tribe, the
Mountain Ute Tribe, and the Navajo Nation. Should
additional data become available from sources with
quality assurance documentation, we will strive to
incorporate them into our analysis.
HenkHaitjema
21) In some cases data sources and limitations were not
fully explained and required reviewer inquiries. While in
most cases an appropriate attempt was made to overcome
data scarcity and uncertainty by offering conservative
(worst case) scenarios, these were not always well
explained.
21) The EPA/ORD team presented a discussion of
similarities and differences among data sets obtained
from various agencies and cited some of the
potential uncertainties associated with these data as
well as gaps in available data. In the final report, we
will pay close attention to highlighting and
explaining important gaps and uncertainties in data
and analyses.
Kirk Nordstrom
22) For the most part, the data that were available were
properly included and appropriate. There is the distinct
possibility that additional data was collected by university
researchers, local stakeholders (such as the Animas
Stakeholders Group), mine owners, the US Geological
22) We have acquired data collected by the U.S.
Geological Survey (USGS) during the GKM plume
event, as well as relevant historic data collected in
the Animas and San Juan Rivers and used it in
analysis. Our analysis has utilized data collected by
18
-------
Reviewer Name
Reviewer Comment
EPA Response
Survey (USGS), The Bureau of Land Management
(BLM), and the US Forest Service that has not yet been
discovered. For example, I am aware that some data was
collected by the USGS which has not been included in the
compilation and the presentations. These additional data
sources, which included USGS data given to Steve Way
of the EPA should be found and included if useful for the
modeling.
23) It would have been helpful for me to have the team
include chemical analyses of just a few waters samples
such as GKM effluent in addition to the samples that
defined the tail end of the plume. Then I could do some
quick calculations to both confirm what the team had
calculated and to see if there are any additional
calculations that might need to be considered. The
reviewers only saw a graph of a limited number of
constituents.
24) Several of the plots were log plots that gave a strange
symmetry to the data. I know that in many cases there is
such a large range of values that a log plot is necessary
but not in all cases. Log plots often make the data look
better than it really is. I would suggest that some plots
could be divided into 2 or 3 linear plots for better
visualization.
EPA, states, tribes, and other organizations who
have followed documented quality control
procedures. Should additional data become available
we will to attempt to incorporate it into our analyses.
23) The GKM effluent concentration data will be
included in the final report or its appendices along
with an explanation of which values were selected to
represent the plume and why.
24) The EPA/ORD team uses log-scaled plots to
display data during plume passage that typically
spanned 3 orders of magnitude and sometimes
ranged over 5 to 6 orders of magnitude over the
length of the affected rivers. In the final report we
will design graphical presentation of results to use
the minimal scales possible with the least distortion.
19
-------
Reviewer Name
Reviewer Comment
EPA Response
William
Stubblefield
25) The scope and types of available data were adequately
described and the limitations of the available data were
also discussed. Obviously, there were limitations in the
available data and in some cases key parameters that
would have been useful for interpreting data were not
available (e.g., dissolved organic carbon). The staff doing
the analyses had to "make do" with the extant data and
they seemed to do an adequate job with what was
provided.
26) In some cases, questions were raised regarding the
potential availability of data from other non-EPA sources
that might exist. EPA is encouraged to seek out and
obtain all potential data that would be useful in
interpreting the extant data. Potential data sources that
should be examined include the USGS and State
Department of Environmental Quality and/or
Departments of Fish and Wildlife. In addition, it is
anticipated that there may be data held by researchers at
local Universities and at various Native American
organizations.
25) Although there is considerable data on metals in
water and bed sediments, there are very little data on
dissolved organic carbon during the plume, as well
as some other parameters that would have facilitated
geochemical analysis. However, we feel that a
robust analysis addressing the project objectives can
be accomplished with the available data.
26) The EPA/ORD team has acquired data from
EPA, the states of Colorado, New Mexico, and Utah,
and the Southern Ute Indian Tribe, the Ute Mountain
Tribe, and the Navajo Nation. We have also
acquired data collected by the US Geological Survey
(USGS) during the GKM plume event, as well as
their historic data. We continue to acquire data and
reports by various entities as they are released.
Should additional data become available we will
strive to incorporate them into our analysis.
20
-------
Question 2
Was the overall integration process of the various analyses conducted in a way that
provided meaningful results and conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
27) In general the overall integration process of the various
analyses at NERL was conducted in a way that provided
some meaningful results and conclusions. However, the
integration process outside of NERL appears to be a
significant barrier to deriving more meaningful results. The
lack of consistency in the data between different
organizations, data gaps for some important analyses (such
as pH and conductivity), and different detection limits and
analytes even for the EPA labs, all make the overall
integration appear weaker. In addition the apparent lack of
integration between ORD NERL, other ORD labs, the
regions, and other agencies in terms of response and future
monitoring and modeling needs, limits the provision of
meaningful results and conclusions. With regard to the
presentations, it probably would have been more helpful to
present the empirical results before the WASP modeling.
27) The EPA/ORD team used publically
available data to address our primary research
objectives of characterizing the source, transport
and fate of metals in the GKM plume to a
reasonable level. EPA samples were collected
and laboratory tested following Quality
Assurance Project Plans and Standard Operating
Procedures in place in each Region. Field
sampling included pH and conductivity. We also
used state and tribal data collected following
documented QA/QC procedures. Data among
organizations was generally very consistent.
Additional data obtained from various sources,
including sondes, added insights into plume
movement. Given that all of these entities were
responding to an emergency event, all of the
groups together managed to compile a robust
collection of data that we feel provided a strong
and coherent analysis of the GKM event at its
origin and as it traveled 600 kilometers
downstream. Often the mix of data sources
enabled the gaps in one place or time to be filled
by nearby measurements. Uncertainties are
present, but the overall story was internally
consistent and supported independently by the
modeling that we deployed. Differences such as
detection limits affected the ability to draw
21
-------
Question 2
Was the overall integration process of the various analyses conducted in a way that
provided meaningful results and conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
conclusions about low concentrations of metals,
raising uncertainty as to when and if the metals
returned to pre-event levels, but did not deter
detection of the GKM plume as it traveled.
Differences in detection limits and lack of
consistency in data sets were more significant in
historic data that has been collected for many
different reasons. Again, this increases
uncertainty in determining if the GKM plume
has completely left the system or when the river
concentrations will return to pre-event levels.
Charles Fitts
28) I understand that there has been a pressing timeline for
pulling these studies together and that we are looking at
first drafts, which I think is the proper stage for having a
review that allows time for revision. I expect more effort
will go into integration, peer editing, and polishing, which
the entire study could benefit from.
29) It would help to expand the overview section so that it
explains clearly how each of these parts contribute to
achieving the project's goals and describes to what extent
each part depends on results from other parts. For example,
the same analysis of source mass shows up in both the
empirical and WASP sections.
30) Some portions of the work could benefit from
additional review and input by additional experts within
28) The EPA/ORD team continues to improve
the presentation of results and has worked to
improve integration between software modeling
systems and empirical analysis of observed data
utilizing the suggestions and feedback of the
peer reviewers. These improvements will be
implemented in the final report.
29). The EPA/ORD team will include a
"roadmap" of the analysis as to how data was
used and how modeling approaches (empirical
and analytical) interacted to address the research
questions.
22
-------
Question 2
Was the overall integration process of the various analyses conducted in a way that
provided meaningful results and conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
EPA. Although I am not expert in this area, it seemed that
the bioaccumulation study could use such review, as the
reviewers indicated that it may need to consider alternate
methods that are not based on factors that assume a ratio of
river concentration to tissue concentration.
30) The EPA/ORD team agrees and will include
effects-based expertise.
HenkHaitjema
31) There were some limited connections between the
presentations, particularly between the presentation
"Empirical Analysis of Metal Loads & Water Quality
Trends Based on Observed Data" by Dr. Kate Sullivan and
Dr. Mike Cyterski and the WASP modeling. However,
there was no clear overarching structure in which the
various presentations had a clear place. Consequently, the
results and conclusions from the individual studies could
not easily be related to each other. That said, I recognize
that this review was conducted before all studies were fully
completed and documented (written up in a report) and as a
result the integration could not yet have happened. I believe
that the timing of this review, prior to producing a final
document, is very beneficial for an optimal impact of the
review process. Thus the lack of integration observed is not
to be interpreted as a critique on this research effort!
31) The connection between the empirical
analysis of water quality concentrations and
plume movement and the WASP modeling has
been strengthened in response to the peer
reviewer's comments. This includes stronger
calibration of WASP against the empirically
derived load estimates to calibrate particle
settling. Integration of analysis will be stressed
in the final report. In addition, the modeling will
be used more prominently to animate findings to
facilitate presentation of findings to the public.
23
-------
Question 2
Was the overall integration process of the various analyses conducted in a way that
provided meaningful results and conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Kirk Nordstrom
32) The presentation of the various analyses could have
been conducted in a logical sequence. The Empirical
Analysis should always go before any modeling efforts
based on the observations. Most people would want to see
the data first and foremost. It is also better to get a feeling
for the data to see what types of modeling approaches are
reasonable and which ones aren't. Modeling is usually used
to fill in data gaps, to gain more insight into the processes
that might explain the data, and to explore possible
scenarios to evaluate their consequences. So the data should
come first and then the modeling results. Otherwise the
sequence with Geochemistry, followed by WASP
modeling, Bioaccumulation, and ending with Groundwater
seemed appropriate.
33) One aspect that was problematic is that some of the
results and the presentations changed several times. That is,
we received one copy of PowerPoints by cyberspace before
the meeting. At the meeting we received a paper copy of the
PowerPoints in a binder where some things had been
changed and then when people gave presentations they
sometimes had made another update and handed that out to
us separately. That tells me that the Team was not quite
ready and were still finessing their results. It would have
been more appropriate to wait another week or two to make
sure there were no important changes before presenting to
the reviewers. Last minute modifications are not helpful for
a review meeting.
32) The EPA/ORD team will change the
sequence of presentation of the multiple
elements of the analysis in presentations and
reports to reflect this comment and will more
carefully orchestrate the description of data
concordant with modeling discussions in the
final report.
33) The analysis was ongoing at the time of the
review which was scheduled to evaluate progress
to date. Input from members of the panel helped
further refine and improve analyses. The final
written report will fully describe updated
analyses and will undergo additional external
peer review.
24
-------
Question 2
Was the overall integration process of the various analyses conducted in a way that
provided meaningful results and conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
William
Stubblefield
34) Is not entirely clear what is meant by the "overall
integration process" of the various analyses. For example,
some of the reported metals data are presented on the basis
of "total metals." This is a fairly nonstandard approach for
presenting metals data especially if one of the objectives of
the evaluation is to assess potential impacts to exposed
aquatic organisms. The array of metals present in the Gold
King Mine AMD will have vastly different toxic potencies
and will be present in the AMD at greatly different
concentrations (ppm to ppb). To conduct an appropriate
evaluation of potential effects to exposed organisms, one
needs to consider the exposure to the individual metals. It
might be better if evaluations were conducted on a few
different metals representing a range of toxicities,
proportional presence in the AMD, and environmental fate
processes. Evaluating metals such as iron, aluminum,
copper, and zinc would cover a range of toxicity profiles
and presence in the AMD.
34) In the interest of maintaining a reasonable
number of individual figures, the EPA/ORD
team sometimes presents summed metals to
demonstrate the general patterns of metals
transport by concentration or mass that were
observed with the plume. The peer review panel
was generally dissatisfied with that
simplification as individual metals are of interest
biologically and geochemically and probably
behaved differently within the system and with
differing levels of importance. The final report
will describe general plume movement with
summed metals as an overall description of
movement but will analyze individual metals as
well.
25
-------
Question 3
When looking at the full project, are there errors or gaps in the integration process
that could have affected the overall analyses and/or the conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
35) Please see comment on question 2 above.
35) See response (27)
Charles Fitts
36) Since the study focused mostly on total metal concentrations,
it is possible that it overlooked behaviors of specific species of
metals that would be important in subsequent risk assessments and
monitoring plans. I also mention this in the fate and transport
section, and suggest analyzing the fate and transport of a few
metals that are likely to pose risk and may be representative of
groups of similar metals.
37) Most of the concentration data we saw in the presentations
was from water samples. However, slides 6 and 20 of the
overview alluded to over 320 bed sediment samples. Presentation
of the sediment data was limited, so if there is more to that story,
perhaps more should be presented.
36) The EPA/ORD team will emphasize analysis of
individual metals in the final report.
37) The presentation of data included a representative
number of sediment related analyses. Future
presentations and reports will include a full analysis
of bed sediment data.
HenkHaitjema
38) While the "Overview" presentation offered a "Summary of
Findings" (slide 25) that I found relevant and important, there was
no overarching presentation that put the various studies together to
substantiate these final conclusions. What is needed in addition to
the work presented to the reviewers is a document with a clear
statement of purpose and explanation of the motivations for the
various studies. That same document then must also have a
concluding section in which these studies are referenced, and the
conclusions integrated into an overall set of conclusions and,
where appropriate, recommendations. I did not observe
fundamental flaws in the studies that negatively affected the
conclusions presented.
38) The EPA/ORD team appreciates the reviewer's
comments regarding method of presentation. The
final report will be structured to address the
recommendations of the reviewer.
Kirk Nordstrom
39) The integration process could have been improved by better
communication between the Geochemical Analysis and the
26
-------
Question 3
When looking at the full project, are there errors or gaps in the integration process
that could have affected the overall analyses and/or the conclusions? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Empirical Analysis groups. These 2 sections are very closely
aligned and have clear overlaps on the source term composition.
More discussion was probably needed between these groups to
have a better consensus on how to characterize the source term. It
seems to me that when writing up the final report these 2 sections
might be merged into one. Alternatively, writers should make
clear what deserves to be called geochemical analyses and what is
empirical. Whenever geochemical modeling is involved it would
seem necessary to call it a geochemical analysis, however, mass
balances is also considered geochemical modeling. Very often
some geochemical reactions need to be assumed or modeled for
the mass balances to make sense. Hence, these two sections should
probably be merged.
39) The EPA/ORD team will improve description of
the coordination between geochemical analysis with
the various modeling and empirical approaches used
and integrate discussions of these topics in the final
report.
William
Stubblefield
40) One of the stated objectives of the effort was "Identify the
potential for water quality impacts, including municipal wells, and
implications for future monitoring priorities near-term and long-
term." It is not clear how this objective was going to be met. Few
"exposure" concentrations were provided as a result of the
Agencies analysis and little to no indications of how "impacts"
were going to be assessed were discussed.
40) We have added more presentation of the
potential exposure to metals for various water uses
during and following the GKM event relative to
water quality criteria. These will be included in the
final report.
27
-------
Question 4
Were the overall conclusions that were drawn from these analyses appropriate and
scientifically defensible based on the analysis? Why or why not?
Reviewer Name
Brian Caruso
Charles Fitts
Reviewer Comment
41) The overall conclusions drawn from these analyses generally
seem appropriate, but this is somewhat difficult to determine due
to the lack of clarity in the goals and objectives of the research. In
addition, the conclusions are not entirely scientifically defensible
based on the analysis. The primary reasons for this are generally
discussed in the overall impressions above and include:
lack of clarity of goals and obj ectives to reflect critical
information needs
lack of use EPA national criteria or standards for metals
for drinking water and aquatic life as an initial screening
of risk to select potential contaminants of concern for
more detailed analysis and as indicators (instead of
primarily evaluating total metals)
inadequate use of other existing data and information
from previous investigations to evaluate and help confirm
background levels for comparison
lack of inclusion of additional data and information for
better reactive transport modeling, concentration and load
calibration, and validation for WASP
very limited evaluation and presentation of uncertainty
and sensitivity analysis of results
lack of identification of data gaps in the analysis and for
future modeling
42) I think the conclusions presented in overview slide 25 are
generally sound and on-target. It think that the 4th bullet point
about most of the metals being deposited in the Animas
streambed could be more specific. The presentation could point
EPA Response
41) The EPA/ORD team presented the peer review
team details of data analysis and modeling of
volumes and metals concentrations at the GKM
source, the timing and concentrations of metals at
the plume traveled through the Animas and San Juan
River systems, the fate of metals mass in surface
water and sediments, including possible transport
through alluvial sediments into groundwater wells.
Quantifying these aspects of the GKM release was
the main objective of the project as described to the
peer review team in each individual unit of analysis.
As recommended, we will include additional
analysis of historical data, increase calibration of
WASP with observed data, and will address
sensitivity and uncertainties associated with data and
modeling. These will be included in the final report.
42) This reviewer interpreted the information
correctly as to where most of the deposition
occurred. The EPA/ORD team will ensure that these
locations are highlighted as the areas with greatest
28
-------
Question 4
Were the overall conclusions that were drawn from these analyses appropriate and
scientifically defensible based on the analysis? Why or why not?
Reviewer Name
Reviewer Comment
EPA Response
out the specific stretches of the Animas River that received the
greatest mass of deposition (RK 13-16 and RK 64-96, as
discussed in question 7). There is no bullet point about the impact
on wells located near the river. I think there should be an
additional point made about the potential for impact in wells
close enough to the river, but that sampling data showed only
well 35m66km with a noticeable plume signal, which was at
levels that did not pose any significant risk.
deposition in the final report. We will also include
discussion of the potential impacts on wells in these
areas and include findings on groundwater potential
exposure in summary of findings in the final report.
Henk Haitjema
43) I believe they were, but as outlined in my response to various
questions below, additional work and better documentation are
needed.
43) The EPA/ORD team agrees and will address in
the final report.
Kirk Nordstrom
44) Not entirely. (1) The geochemical analysis used some flawed
assumptions to estimate the GKM effluent composition (see
below), (2) alternative approaches to the GKM effluent
composition were not considered (see below), (3) sensitivity
analyses need to be employed for many of the analyses and
modeling with a propagated range of uncertainty; this approach
would result in upper and lower bounds for the plume at several
locations downstream, and (4) I have a difficulty in seeing any
scientifically defensible conclusions coming out of the
bioaccumulation study - the lack offish kills and the caged fish
study are much more appropriate to address fish toxicity for such
a short transient event than the attempt at modeling that was
presented.
44) Items (1), (2) will be discussed in later
responses following Dr. Nordstrom's suggestion to
see his responses below where he provides more
detail.
(3) The EPA/ORD team will include discussion of
uncertainty and sensitivity to assumptions in the
final report.
(4) We view some treatment of bioaccumulation
modeling as helpful in understanding the potential
uptake of metals during relative rapid movement of
metals through their habitat and help to corroborate
what was observed in fish within the system and to
assess potential for dietary uptake. The revised work
will be presented in the final report that will also
receive an external peer review before publication.
29
-------
Question 4
Were the overall conclusions that were drawn from these analyses appropriate and
scientifically defensible based on the analysis? Why or why not?
Reviewer Name
Reviewer Comment
EPA Response
45) A variety of conclusions were provided in a number of the
presentations; however, for the purposes of this response, we are
assuming that the "summary of key findings" from the overview
presentation captures the "overall" conclusions. For the most part
these findings were supported by the data provided in the
presentations. However, in some cases it is difficult to point
specifically to the data that support a given conclusion. This is in
part due to the sheer volume of data and the way that the
presentations were organized based on the available time for
presentation. It is anticipated that a detailed report outlining the
analysis that was conducted would provide an opportunity to
present an analysis in greater detail. For example, providing
metal specific data rather than "total metals data" would provide
greater support for the conclusions.
45). The final report will present data organized in a
manner that facilitates integration of findings. This
will include mapping between data and analyses
used for or generated by them. Key individual
metals will be showcased.
30
-------
Part 2: Fate and Transport
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
46) The research makes an attempt to characterize the metals concentrations and
loads produced at the Gold King Mine spill. However, it is extremely surprising
and unfortunate that EPA collected no samples from the release itself until what
appears to be a substantial time period after the release. In addition, no samples
were collected at the mouth of Cement Creek (CC) until about 4 hours after the
release and after the release/plume had passed. The volume of the release was
estimated by the USGS based on the change in the hydrograph at the CC mouth.
Four samples were collected at the adit release up to about September 23, 2015. It
was not made clear when the first sample at the adit was collected, but appears to
be at least many hours to a day after the release. One of these samples collected by
EPA was selected to characterize the release and use in subsequent calculations
and modeling. It was stated that this was selected because it was the most
comprehensive analysis. However, it is not known or made clear why the other
samples were not analyzed the same way. The samples are presented on a log
graph for most metals, so the variability of the results is not entirely clear. The
variability and uncertainty of these adit release results should be analyzed and
presented in more detail, and perhaps a mean or median over this time period
should be used instead of just one sample. Also, it is not clear if any other samples
from inside the adit itself, or from the ongoing drainage, had been collected and
analyzed previously, prior to the release. If so, these should be compared to what
was observed in the release.
46) The first samples of mine effluent
were collected by Colorado DPHE
within 2 days of the event and 3 more
measurements were made by CO and
EPA over the next 6-weeks. These
data were shown to the peer review
team and will be provided in the final
report. In the final report, we will
provide detailed description of these
data and analysis used to select a
value to represent the effluent
concentration, including a discussion
of uncertainty.
Charles Fitts
47) The data were mostly presented as total metals and did an adequate job of
portraying the distribution of total metals. The presentation seldom presented data
on subgroups of metals or individual metals. It might be instructive to look at
empirical data for a few individual metals of interest, selected because of their
importance in terms of risk and their characteristic behavior representative of a
47) The EPA/ORD team will
continue to strive to find
parsimonious ways to present data as
appropriately as we can. The final
report will describe general plume
31
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
group of similar metals (e.g. one metal that precipitates at a low pH range and
another that precipitates at a higher pH range). Since subsequent studies will be
examining risk and monitoring plans that aim to minimize risk, the metals chosen
for individual analysis should include ones that are most likely to pose risk.
movement with summed metals as an
overall description of movement.
Individual metals concentrations,
mass loads, and exposure analysis
will be included in the final report.
HenkHaitjema
48) As explained below the total metals load leaving Cement Creek were probably
underestimated. However, this was recognized in the analyses presented to the
reviewers and could not have been avoided in lieu of the lack of more pertinent
sampling (sampling of the peak of the plume in Cement Creek).
48) The EPA/ORD team and the peer
reviewers discussed mass estimates
from Cement Creek at great length.
There are important uncertainties that
would not likely be bridged with data
collected after the fact given the
unique conditions during the event.
Kirk Nordstrom
49) Characterizing the composition and load of the Gold King Mine spill is
problematic. No samples of the mine effluent were taken during the spill event.
Samples were taken some days later. When the plume hit the first gage at Cement
Creek (at the mouth), samples for chemical analysis were taken well after the peak
of the plume had passed. Furthermore, the first 2 samples at the gage were
incomplete (no pH, conductivity, or sulfate determinations). In addition, when the
plume hit the Cement Creek gage it had picked up additional sediment and
dissolved substances that were not part of the original mine pool discharge.
Consequently, it makes sense to consider the source water as the plume that was
recorded in the Cement Creek gage right before it entered the Animas River. It is
still a problem characterizing the water composition at the peak of the Cement
Creek discharge because the first sample collected for analysis at the gage was
about 5 hours after the spill began and contained only about 20% of the Gold King
effluent as well as missing some critical parameters. I think the ART did important
calculations to estimate the water composition at the Cement Creek gage peak
flow from the GKM release and I shall suggest additional considerations.
49-50) The EPA/ORD team agrees.
The metals concentrations and
therefore mass at the peak of the
GKM plume in Cement was likely
much greater than first sampled at
16:00 hr after most of the plume had
passed. Our methods to reconstruct
the plume that were shared with the
panel are generally described by this
reviewer in (50) and (51).
32
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
50) The Team did a straightforward conservative calculation assuming straight
mixing of GKM effluent with upper Cement Creek water with no reaction. This
result would normally give a bounding limit to the chemical composition of the
plume. But which limit? High or low? If there is a reaction in progress, is that
increasing or decreasing metal concentrations? Both are possible. Oxidation and
precipitation of iron would tend to remove metals. Dissolution of soluble salts
from the eroded waste piles and Cement Creek would increase metals. Erosion of
fine clays might provide more surfaces for metal sorption and partitioning from
dissolved to the solid phase. From my experience with weathering of mine tailings
and waste rock during storm events, there is a brief and sudden increase in
dissolved metals during the early rise of the discharge and then a decrease from
dilution. In this instance, dilution is with GKM release water and upper Cement
Creek flow because it is not a rainstorm event. But there is still likely to be a
sudden increase early in the plume movement and then a drop to the
concentrations of the GKM effluent for the remaining majority of the plume
release followed by decrease to Cement Creek baseline once the GKM plume has
passed. This early spike in concentrations would be from the addition of soluble
salts and films of concentrated acid mine water contained within the tailings pile
downstream and separate from the effluent composition released from the mine. I
would anticipate sorption processes to be largely ineffective at this pH (~3) and
with higher than normal metal concentrations. The plume is moving too fast for
much oxidation and precipitation of iron. Hence, I would argue that the total
plume load would be greater than that expected from just the analyses of the GKM
effluent in both dissolved and fine particulate matter combined with the estimated
discharge. Further, I would argue that the first measured concentrations at the gage
on Cement Creek should be close to conservative mixing (20% of GKM and 80%
upper Cement Creek water) but that the dissolved concentrations were higher
50)
We will revisit how to characterize
the dissolved fraction of the metals
concentrations at the peak of the
plume in Cement Creek. The final
approach that we use to characterize
the metals in Cement Creek will be
fully described in the final report.
33
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
during the first 1A hour of the GKM release. How much higher is very difficult to
say so this calculation would be a lower bound that can be compared to another
estimate. It can also be compared to a loading calculation that takes a constant
composition GKM release as a lower limit after mixing with Cement Creek
baseline water. This constant composition chosen by the ART was the August 15
sample because it was the first complete analysis of the mine effluent after the
plume had passed. There were 3 other samples that I would say could be used as
well from other time periods. Although pH, sulfate concentration, and
conductivity data were sometimes missing, it is possible to reconstruct these by
optimizing pH and sulfate concentrations using charge balance for pH (using the
PHREEQC program) and conductivity balance (using either PHREEQC or
WATEQ4F although WATEQ4F would be preferable because it is more reliable
for acid mine waters).
51) With regard to estimating the composition of the Gold King effluent water
during the spill, the explanation could have been clearer, especially since this
composition is critical to the entire interpretation of downstream fate and
transport. Unfortunately, the data available is sparse and incomplete which adds to
the confusion. As I understand it, there are two key sets of data: (1) direct analyses
of the Gold King effluent but collected after most of the spill had occurred with
dates of 8-07-15 and 8-11-15 collected by CDPHE and dates of 8-15-15 and 9-21-
15 collected by the EPA and (2) Cement Creek samples collected during the tail
end of the plume movement (first sample was collected about 5 hours after the
spill began). The CDPHE samples are missing critical data such as pH,
temperature, conductivity, iron and sulfate concentrations. The Cement Creek
samples are Gold King effluent mixed with 80% or more of upper Cement Creek
water, possibly mixed with some dissolved soluble salts, eroded sediments, and
their pore waters. The GKM effluent composition had to be estimated from these
51) Regarding alunite: based on
review of the literature, important
details of alunite chemistry are still in
need of resolution. Also, based on
discussions during peer review, the
EPA/ORD team has performed
kinetic calculations on alunite
dissolution rate with data recently
published in the literature; these
calculations also argue against alunite
dissolution accounting for the high
aluminum reported in the Cement
Creek sample. Because of these
uncertainties, the EPA/ORD team
agrees that alunite chemistry cannot
be invoked to estimate load limits
with any confidence and load and
source estimates based on
equilibrium chemistry driven by
alunite will not be used in the final
report.
34
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
limited pieces of data. The approach taken was to use the Cement Creek USGS
gage data to determine the proportion of that water containing GKM effluent.
Then unmix the water assuming conservative mixing. Then most of the
concentrations were increased by an amount that was estimated by assuming that
alunite saturation equilibrium was achieved in the GKM effluent and increasing
the aluminum concentration accordingly. Alunite saturation equilibrium was
indicated in a paper by Eary (1999) and this is the first time I have heard of
making this assumption to estimate a mine water composition. The question is
whether this assumption is reasonable and whether there are other, more
reasonable approaches. Alunite is a relatively insoluble mineral which is slow to
dissolve and precipitate unless the temperature is increased substantially above
ambient.
I have read the Eary (1999) paper and the case made for alunite solubility
equilibrium at low pH is extremely speculative. I say that because the plots that
Eary showed (1) had considerable scatter, (2) were not done the normal way with
the log of the activity of the free aluminum ion vs pH - he used dissolved
aluminum concentrations vs. pH which doesn't really tell you much and cannot be
directly compared to solubility of alunite, and (3) he doesn't show saturation
indices for alunite as he does for gypsum, fluorite, and other carbonate and sulfate
minerals. Further, he was looking at a pit lake which can be different than
underground mine effluent. Not to mention that there are a range of
thermodynamic properties for alunite so we really don't know how the solubility
might change with solid solution substitution, particle size and crystallinity, and
uncertainty in the thermodynamic properties. I am sure that alunite does reach
equilibrium solubility in some environments but I would be very hesitant to apply
it for this situation. Hence, I would discourage using this type of modeling
approach to correct the mine effluent chemistry to the original composition.
52) We will refer to the QAPPs used
by EPA regions for sample collection
35
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
Instead, I would take the range of composition of the mine effluent water (max
and min as bounding conditions) that was sampled later, correcting pH and sulfate
concentrations as mentioned above, and compare that to the conservative estimate
made from the mixing calculation that the team did from the 1600 hour sample.
Then I would consider a 50% to a 100% increase in concentrations during the first
!/2 hour only of the GKM release to account for washout of the tailings pile for an
upper bound of the loading and concentrations.
52) The characterization of the metals concentrations and the loads begins with the
field collection of water samples and field parameters, followed by laboratory
analyses. The ART did not participate in these activities. There may have been
some QA/QC (quality assurance/quality control) tasks done by individuals in the
team, but, apparently not as a group effort. Consequently, some unexplained
discrepancies occurred in the results presented, such as several elements in which
the total (unfiltered, acidified) concentration is substantially less than the
dissolved (filtered, acidified) concentration. This discrepancy is most apparent
with As, Sb, Pb, Mo, and V in the Cement Creek samples that were used to
estimate the source effluent composition from the mine, which are sometimes
discrepant by an order of magnitude or more and that is far greater than the
analytical error. One way of avoiding these problems is for the team to engage in
conversation with the field collection personnel and with the laboratory and any
QA/QC examiners to determine if there were any sampling problems or analytical
problems that could explain these anomalies. I have seen similar discrepancies
before with metal concentration data from mine-influenced water at Superfund
mine sites and the main problem seemed to be the lack of communication between
those collecting the samples, those analyzing the samples, and those providing
QA/QC. Without knowing field difficulties in collecting samples and whether
there were any modifications of normal procedures (waters should be filtered and
and testing and include this
information in the final report. This
project also operates under an
approved Quality Assurance Project
Plan with provisions for data quality
control.
36
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
acidified immediately on collection; unfiltered samples acidified immediately
except for anion sample) and without knowing if any serious interferences or
possible contamination occurred with the analytical procedures, it becomes
impossible to know how best to interpret the data. The higher dissolved
concentration could be a contamination problem and the lower total value closer to
the truly dissolved value OR the dissolved concentration could be more accurate,
and the total concentration could be a result of the sample being collected in a
different part of the river or an analytical interference. These are important issues
that can affect any attempts at interpreting the results for fate and transport.
53) For this report, everything that can be known about sampling, preservation,
and analytical procedures should be spelled out more. There were probably
different procedures employed by State, Federal, tribal groups and other parties
(for example, were samples sometimes stored for some time before acidification?
Was the same acid used among agencies for acidification? Was acidification done
with the same strength acid and with the same volume per volume of sample or to
the same pH? If samples were filtered, what was the filter pore size? Instead of
providing the EPA method numbers for the analytical method, it would be better
for the reviewers to simply have the actual instrumental technique employed (ICP-
AES or ICP-MS, etc.) which might be more useful when comparing results from
different agencies. Reviewers and stakeholders might want to know the QA/QC
for the data. I recommend a table that lists what samples were collected when, by
whom, whether filtered on site or not, if filtered what pore size was used, whether
acidified on site or later, if later how much later, what and how much acid was
used. A separate table can cover QA/QC data (blanks, spiked recoveries, standard
reference water samples, alternate methods). These tables can be appendices in the
report, but it is essential to include this information because it supports the
credibility and usefulness of the data for modeling and interpretation.
53) The EPA/ORD team will further
research and compile information on
data collections methods for data
used in this study. The final report
will provide summary tables and
links to quality assurance
documentation.
37
-------
Question 5
Were the overall conclusions that were drawn from these analyses appropriate and scientifically defensible based on the analysis? Why
or why not?
Reviewer Name
Reviewer Comment
EPA Response
54) Also, several metal concentrations that were reported are of questionable value
such as cobalt, barium, and beryllium. I know these are easy to determine by ICP-
AES and ICP-MS but if there are no obvious toxicological concerns and the
concentrations are quite low, then that could be stated explicitly. It could also be
stated that certain metals were selected (and others not) for continued description
in the plume movement because of their concentrations and their potential
toxicity.
55) My understanding is that grab samples were collected rather than width-
integrated composite samples. Under the given conditions, it might be that grab
samples were the only ones possible at many of the sites, however, some width-
integrated samples should have been possible or at least near-central-velocity
samples collected. If the team doesn't know what the velocity of river was where
the sample was collected, it could easily affect the results. Some information on
this aspect should be provided in the final report.
54) A number of metals were present
in only very low concentrations and
below detection limits. The final
report will discuss how samples at
detection limit concentrations are
used and what metals are more fully
analyzed and reported and why.
55) There is inherently variability in
both surface water and sediment
samples due to collection techniques
used by different organizations. Many
surface water samples were grab
sampled rather than using depth
integrated techniques. This could add
to variation among samples, although
much of the river system was
sufficiently turbulent to minimize
these errors. The final report will
describe river conditions that would
contribute to mixing assumptions.
William
Stubblefield
56) It is difficult to address this question given the "total metals" approach used in
the analysis of the data. It would seem logical that there are sufficient individual
metals data to permit a "by metal" analysis of exposures. This would be helpful in
addressing the questions associated with potential impacts to organisms and would
allow for better characterization of the fate and transport of individual metals.
56) Information and analyses of
individual metals will be emphasized
in the final report.
38
-------
Question 6
The concentration of metals near the release site in the receiving waters had to be estimated from samples collected after the
much of the plume had passed. Were the estimates of metals concentration at this location appropriately calculated through
scientifically sound methods using available data?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
57) There was a reasonable attempt made to estimate metals
concentrations at this location (adit and or CC?) using scientifically
sound methods based on available data. As stated above, however,
there appears to be many questions and issues with regard to the
analysis methods and assumptions.
58) At the adit release, estimated concentrations and loads were only
based on one sample, whereas the summary statistics, variability, and
uncertainty of the four samples collected over the month and half
after the release should have been better presented and perhaps used
in the analysis. Downstream at the CC mouth, an attempt was made
to back calculate the concentrations and loads during the peak flow,
and to account for dissolved and particulate metals scoured from CC
by the passing flood wave. It is not clear how WASP was used to
calculate the Maximum Total Concentration to aid with this. This
appears to be done outside of WASP as input to the model as a
simple mass balance using the estimated release concentrations,
estimated background upstream CC concentrations and flow, and
downstream measured flow. This mass balance approach seems to be
appropriate. However, the analysis is not clear and background
concentrations in CC appear to have been based on post-plume
concentrations at the mouth, even though there are many pre-release
sample and analysis data available for CC. These previous data could
have, and probably should have been used, or at least collected and
compared to the background estimates used.
57) The data and analysis method of
metals concentrations in the mine will
be fully discussed in the final report.
58) The data and analysis method of
metals concentrations in the mine will
be fully discussed in the final report.
We clarify that we did not use WASP
to calculate the metals concentration
at the peak flow in Cement Creek but
rather we used the mass balance
approach described by this reviewer.
This method will be fully described
in the final report. If background
concentrations in Cement Creek are
needed for the analysis the supporting
data will be fully described.
39
-------
Question 6
The concentration of metals
near the release site in the receiving waters had to be estimated from samples collected after the
much of the plume had passed. Were the estimates of metals concentration at this location appropriately calculated through
scientifically sound methods using available data?
Reviewer Name
Charles Fitts
Reviewer
Comment
59) The explanation of how the plume concentrations were re-
constructed at the CC mouth is not
entirely clear. It is not clear
whether the PHREEQ modeling was needed,
or what value the
WASP modeled concentrations are considering; these are estimates
based on conservative constituents
with no reactive transport.
60) The calculations that lead to the "estimated peak" concentrations
shown in the bar chart of empirical slide 20, the WASP and
Empirical
concentrations
in slide 27, and "Simulated Load" in
WASP slide 13 needs to be explained in more detail. This is critical
since the extrapolation needed at early times strongly affects the
estimated total load in the plume. In the following table, I analyzed
total concentrations (Ct) vs. discharge (Q) for the early time
observations and
the early time simulated concentrations in
Silverton.
Time
12:45
12:45
16:00
19:25
Q
(cms)
3.5
3.5
1.1
0.1
Ct
simulate
d (mg/L)
37000
29557
10500
3000
Ct
observed
(mg/L)
11485
998
Ratio Ct
simul./Q
(mg/L/cms)
10571
8444
9545
30000
Ratio Ct
observ./Q
(mg/L/cm
s)
10441
9980
Source
WASP slide 13
(W13)
Empirical slide
20
W13
W13
EPA Response
59) WASP was not used for this
purpose of reconstructing the
dissolved and colloidal/particulate
concentrations during the GKM
plume at the mouth of Cement Creek.
The methods used will be fully
described in the final report.
60) We have ensured that the initial
value of Ct is consistent across all
analyses and we will clarify the
description of the method in the final
report.
40
-------
Question 6
The concentration of metals near the release site in the receiving waters had to be estimated from samples collected after the
much of the plume had passed. Were the estimates of metals concentration at this location appropriately calculated through
scientifically sound methods using available data?
Reviewer Name
Reviewer Comment
EPA Response
It makes sense that higher stream discharge and velocity would
correlate to higher suspended sediment load and higher total
concentration. Based on observations, it appears that the ratio Ct/Q
was about 10,000 at the earliest observations (blue). Lacking other
evidence, this ratio may be reasonably applied to earlier times, but
there is uncertainty in this extrapolation that should be acknowledged
in the report. The simulated ratio Ct/Q was about 10,000 for both the
12:45 and 16:00 Ct estimates (purple), but about 30,000 for the 19:25
Ct estimate (red). It seems reasonable to keep the 10,000 ratio to
estimate the early time concentrations, which is close to what was
done. I think the bullet item on WASP slide 13 should read "Assume
total concentration (Ct) is proportional to flow." Also the 12:45 and
early time Ct numbers should be made consistent across the study
and consistent with the explanation for how the early concentrations
were extrapolated.
61) Looking at the graph in WASP slide 13 and the 19:25 row in the
above table, the simulated concentrations from about 18:00 onward
are systematically higher than observed, and they are noisy, bouncing
up and down as though the simulated concentrations could only
move in large quantum leaps. This portion of the simulated Ct should
be modified to remove the noise and to better match observed Ct,
even if the impact on simulation results downstream is minor.
62) Since the source mass is critical to all analyses, this deserves
attention. If other concentration data becomes available from other
61) We will examine model output.
Results will be reflected in the final
report.
62) We will incorporate any new
data in the final analysis and report.
41
-------
Question 6
The concentration of metals near the release site in the receiving waters had to be estimated from samples collected after the
much of the plume had passed. Were the estimates of metals concentration at this location appropriately calculated through
scientifically sound methods using available data?
Reviewer Name
Reviewer Comment
EPA Response
sources for the early hours in Cement Creek or GKM, it should be
incorporated in revised source estimates.
HenkHaitjema
63) On slide 20 of the "Empirical Analysis" presentation, two
approaches are mentioned to arrive at the maximum total
concentration (CMAX) in the peak of the plume at 12:45. These are
using WASP for CMAX and PHREEQ for maximum dissolved
concentration. In fact, as I understand it, WASP was not involved in
determining CMAX but a mass balance calculation outside of WASP
was used (see discussion under question 14). I cannot comment on
the PHREEQ method due to unfamiliarity with this code and the
processes it simulates.
64) My overall assessment is that the dissolved concentrations in the
peak are probably fairly well estimated, but that the suspended total
metals concentration in the peak is almost certainly significantly
underestimated. In fact, this is recognized in the current study on
slide 20 with the comment on the graph: "Concentrations at 12:45
peak probably much higher." In summary, the current study does
offer reasonable estimates of peak concentrations and recognizes the
underestimation due to the unknown amounts of suspended materials
in the peak of the plume in Cement Creek.
63) We will clarify and add
additional description of the methods
in the final report.
64) We understand that the peak
concentration as the GKM plume
passed may be higher than the later
16:00 hr sample. The panel offered
insight on how the total and dissolved
fractions could be estimated. The
modified approach that we use will
be fully described in the final report.
Kirk Nordstrom
65) Not entirely. More use should have been made of historical data.
This was mostly addressed above. There is a serious problem with
some of the analyses (e.g. CC06 and GKM13 collected on
8/15/2015) in that many of the total concentrations of metals were
lower than the dissolved concentrations. This can occur from
problems with field sampling and samples that were not filtered and
65) Each EPA Region followed a
Sampling and Analysis Plan when
collecting samples (The sampling and
quality assurance plans can be
accessed at
https://www.epa.gov/goldkingmine/s
42
-------
Question 6
The concentration of metals near the release site in the receiving waters had to be estimated from samples collected after the
much of the plume had passed. Were the estimates of metals concentration at this location appropriately calculated through
scientifically sound methods using available data?
Reviewer Name
Reviewer Comment
EPA Response
acidified on site (which probably did not happen for the earlier
collected samples). Hence, a table summarizing the information on
how water samples were collected and when filtered and acidified, is
crucial to interpreting the results. Also, alkalinities of 5 mg/L are
reported for these samples when the pH is too low for there to be any
detectable alkalinity. This contradiction needs to be resolved.
Further, the acidities are reported but I am not sure they are used or
needed anywhere. There are several different methods for acidity so
the result is very method dependent. If there is a need to report these,
then the method used needs to be documented.
eptember-14-2015-interim-sampling-
plans. Samples were filtered and
acidified on site. The EPA/ORD team
will compile information on data
collection methods for data used in
this study. The final report will
provide summary tables and links to
quality assurance documentation
We will check on the referenced
values of alkalinity. We are not aware
of measured values of acidity. We
reported calculated acidity estimates.
We will explain the methods of
calculation and limitations of these
values in the final report.
William
Stubblefield
66) A number of questions were raised regarding the accuracy of the
estimated metals concentrations in the original AMD release. EPA
staff acknowledged that there was a degree of uncertainty associated
with the estimates and this was reflected in the presentations. It was
recommended by the reviewers that EPA adopt an approach that
characterizes the degree of uncertainty associated with the discharge
estimates and incorporate that into the overall presentation. This
would result in something of a "sensitivity analysis" that would
bound the "best-case" and "worst-case" scenarios.
66) The EPA/ORD team will include
discussion of variability and
sensitivity in future presentations and
the report.
43
-------
Question 7
Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the Animas River?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
67) This comment relates to both questions 7 and 8 since they are
related. The intent of this question is not clear. Is the intent of this for
the post-plume period for all rivers, or just for the post-plume period
in CC and during the plume in the Animas and San Juan rivers? Post-
plume is important for later or ongoing resuspension or dissolution of
metals to the water column during higher flows. Of course during the
plume is also important for a number of reasons. Does 'sediment'
refer to bed sediment or particulates (or colloids) in the water column?
We assume this refers to bed sediment.
68) The methods used for estimating sediment metal concentrations
are not clear. Although the methods used for estimating the dissolved
and sediment/particulate metals load from CC during the release
generally seem appropriate based on the mass balance approach, there
appear to be a number of concerning issues. Any previous data from
other studies on bed sediments in CC are not presented or used. These
could include sediment physical characteristics (particle sizes) and
sediment chemistry and metals concentrations. Similarly, any
previous background metals concentrations in water are not presented.
A significant amount of previous studies by USGS and others in CC
and the Animas have been performed with these types of data.
69) It appears that with the exception of sediment data collected by
an academic researcher (Dr. Williams?), EPA collected no sediment
samples or data during or after the plume to help evaluate sediment
metal concentration estimates or calibrate the WASP modeling.
67) The project analyzed and presented
informations on metals in surface water and
sediment during the GKM plume passage and
in the fall months post-event. Sediment refers
to bed sediment. We refer to the non-
dissolved fraction in the water column as
colloidal/particulate. It may become sediment
if it transfers to the bed.
68) The EPA/ORD team is generally not
focused on the bed sediments of Cement Creek
but we do evaluate metals concentrations in
streambed sediments throughout the Animas
River. The final report will include analysis of
historic data available from USGS and other
studies in this area.
69) The EPA/Regional teams sampled bed
sediments throughout the Animas and San
Juan River systems during and after the GKM
plume, collecting over 300 samples at once or
repeatedly at multiple locations and greatly
expanding on sites that had previously been
measured. The sampling and quality assurance
plans can be accessed at
https://www.epa.gov/goldkingmine/september-
14-2015-interim-sampling-plans. There is also
some pre-event data at a few selected locations
44
-------
Question 7
Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the Animas River?
Reviewer Name
Reviewer Comment
EPA Response
70) It appears that sediment (bed) metals concentrations were not
estimated except for those estimated with the WASP model. There are
issues with these estimates, as discussed in comments below.
in the Animas River headwaters streams
(several of which coincide with USGS
sampling sites). These data are used to analyze
deposition from the Gold King Mine plume
and to calibrate WASP. The use of this data
will be fully described in the final report.
70) Observed bed concentrations were
analyzed and discussed in the "Empirical"
section of the presentations and in the WASP
section of the presentation. Metals mass
(rather than concentrations) were emphasized
in the peer review. The final report will fully
discuss bed sediment concentrations and mass.
Charles Fitts
71) The presentation about measured sediment concentrations was
brief: slides 44-48 of the empirical section. There are inherent
difficulties in distinguishing plume-event sediment from other
sediment, and in concentration variations with sample location.
Therefore, it may be difficult to conclude much from sediment
concentration data. There did appear to be declining trends in
sediment concentrations after the plume passed the lower Animas
(lower two graphs of empirical slide 47). There were WASP
simulations that indicated where sediments from this event were likely
to have deposited (WASP slides 25-27), but as I say elsewhere, these
WASP results should be viewed as qualitative, not quantitative.
71) The EPA/ORD team agrees that it will be
difficult to separate the effects of the GKM
plume relative to pre-existing AMD
contamination of bed sediments, except to
look for higher concentrations post event
relative to pre-event. WASP predictions of
deposition zones are corroborated with
observed samples in empirical analysis. These
comparisons will be fully discussed in the final
report.
72) We agree that the mass analysis derived
empirically at individual locations based on
45
-------
Question 7
Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the Animas River?
Reviewer Name
Reviewer Comment
EPA Response
72) A better approach to estimating where plume sediment was
deposited would be to examine time-series of total metal
concentrations at gages along the Animas River. Estimate the total
mass of metals passing a point in the river by numerically integrating
Ct*Q data through time, like appears to be shown in slides 30, 34, 35,
and 44 of the empirical presentation. Mass changes from one station
to the next one downstream could be due to deposition or mass added
or subtracted at tributaries or diversions. From the graph in the lower
right of slide 44, it appears that most metals deposited just below
Silverton (~RK 13-16) and above Durango (~RK 64-96). This
empirical approach to estimating deposition trends has a much
stronger basis than the deposition results shown in WASP slides 25-
27. This analysis should be expanded and highlighted in the empirical
presentation and overview, and the WASP analysis section should
compare its results to the empirical analysis estimates of deposition.
73) The relatively flat stretch of the Animas River below the
confluence with Cement Creek (~RK 13-16) is an area where a
significant fraction of plume suspended sediment probably was
deposited. Average plume flow velocity would have dropped
dramatically beyond the confluence due to the gentler gradient and
wider channel, and the abrupt increase in pH would have promoted
precipitation and sorption. This stretch of the Animas has alternating
riffles and pools and the larger pools would have been particularly
ripe for deposition. The image below shows one >100m pool in the
Animas channel in this area.
flow and measured concentration is the
strongest evidence for how much and where
metals mass deposited and this is the primary
method used to quantify deposition. We have
also used those calculations as part of the
strong calibration of WASP similar to the
recommendations. Methods and results will
be fully discussed in the final report.
73) We agree with the comment regarding the
likelihood of deposition in the reach of the
Animas River between the confluence with
Cement Creek and before the river goes into
the canyon below Silverton. There are other
likely deposition zones as well. These will be
discussed in the final report.
74) We agree that WASP which moves
downstream at an average velocity calibrated
to USGS at-a-station hydraulic geometry
would not capture velocity variability at the
riffle/pool scale. Although our subsequent
calibrations of WASP have improved its
deposition and erosion performance, we note
that the results shared with the peer reviewers
did show a large decline in concentration in
the reach described at left and again at the exit
point of the canyon and that future erosion
46
-------
Question 7
Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the Animas River?
Reviewer Name
Reviewer Comment
EPA Response
74) The WASP modeling was too large-scale and homogenized to
capture the local differences in velocity between riffles and pools.
WASP slides 25-27 indicated minor deposition in this area, but
empirical slide 44 indicates deposition of about 40% of the metals
mass from Cement Creek in this stretch. I suspect significant plume
mass was deposited in these pools and some will move downstream
during subsequent high-discharge events.
75) Note: a typo in empirical slide 38:
acidity" or "Low pH".
ond
line should say "High
76) In empirical slide 39, explain what blue dots are in lower right
plot.
could occur at both points. We agree that
deposited material from the GKM plume will
likely be entrained with bed sediments in
future flow events. WASP cannot capture
localized effects such as pools, since each
WASP segment behaves as a well-mixed
reactor. Our settling rates, however, are based
on the empirical reconstruction of the plume
from observed concentrations, so we capture
any net loss due to these locations as a reach
average. Model simulations will not be able to
capture mobilization of these localized effects,
however. We will discuss these uncertainties
and possibilities in future presentations and the
report.
75) We have corrected the typo.
76) The triangles are measured pH values and
the blue dots are calculated at specific
locations based on the trend line shown in the
regression. We will explain this in future
presentations and the report.
47
-------
Question 7
Were the data analyzed and visualized properly in regards to sediment metal
concentrations in the post-plume period in Cement Creek and the Animas River?
Reviewer Name
Reviewer Comment
EPA Response
77) The graphs in empirical slide 48 need axis labels and a better
explanation.
78) In empirical slide 50, it would provide helpful perspective to show
estimated total metal transport during a typical spring runoff season,
in addition to the estimates for the late August storm.
77) The axis labels with descriptions will be
included in the final report.
78) The final report will include analyses of
the types described.
Henk Haitjema
79) Generally they were, although other reviewers were often critical
of the lumping of metals into a total metals load or concentration.
79) The final report will provide data on key
metals individually.
Kirk Nordstrom
80) It must be stated much more clearly that the sediment load is a
mix of (1) clays eroded from the tailings pile during GKM release, (2)
clays (mostly Fe and Al oxyhydroxides) formed during oxidation and
mixing with downstream transport, and (3) clays eroded from Cement
Creek during turbulent mixing of the GKM plume. Further, these
sediments should be compared to suspended or clay sediments that
have been determined earlier in USGS studies to see what the
chemical differences are and how much they can be related to the
actual GKM release.
80) The EPA/ORD team will describe the
likely mix of sediments entrained between the
mine entrance and the bottom of Cement
Creek as suggested. We will characterize
sediment in the Animas River during or after
the GKM release where data are available.
William
Stubblefield
81) As previously discussed, very little information regarding bed
sediment metal concentrations were provided. Currently, the state-of-
the-science for evaluating metal concentrations in sediments and the
potential impacts on sediment dwelling organisms requires
information about the acid volatile sulfide content of the sediment and
the simultaneously extracted metal concentrations of other metals
present in the sediment. It did not appear that this information was
available for the sites downstream of the Gold King Mine.
81) The EPA/ORD team is not aware of any
publically available data as suggested that was
collected during or after the GKM release. We
will report metals concentrations determined
using the standard methodologies employed by
EPA and others.
48
-------
Question 8
Were the data analyzed and visualized properly in regards to sediment metal concentrations in the post-plume period in
Cement Creek and the San Juan River after receiving mine contaminated water from the Animas River?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
82) Please see comment on question 7 above.
82) See response to comments (67)-
(69)
Charles Fitts
83) I assume the question refers to the San Juan River, not
Cement Creek. By the time the plume reached the San Juan
River, the muted plume signal was hard to detect and sediment
concentrations in the San Juan did not show a discernable plume
signal. The data presented on this subject was brief (empirical
slide 46). Hopefully the accompanying text, when written, will
clearly explain the origin of the data and conclusions drawn
from this slide.
83) The final report will describe the
sediment data and conclusions clearly
and will present analyses of the plume
in the San Juan River.
84) Future presentations and reports
will provide data on key metals
individually.
HenkHaitjema
84) Generally they were, although other reviewers were often
critical of the lumping of metals into a total metals load or
concentration.
William
Stubblefield
85) Very little information was provided regarding sediment
concentrations in the post-plume period in waters downstream
from the GKM and Cement Creek. Information regarding
individual metals would be helpful, however it is recognized that
it will be difficult to attribute specific metals concentrations to
the GKM incident given the ongoing contamination that exists in
the area as a result of other operations and abandoned mines.
85) The final report will provide more
information on bed sediment
concentrations of metals collected
throughout the Animas and San Juan
Rivers.
49
-------
Part 3: Geochemistry
Question 9
Were the geochemical principles to characterize transport and fate of acid mine drainage
appropriately applied and interpreted? Please explain.
Reviewer Name
Charles Fitts
Kirk Nordstrom
Reviewer Comment
86) I am less of an expert in geochemistry, so my comments in this section
are limited. I found the discussion of the American tunnel plugging and
rising groundwater levels (slide 17) quite interesting from a hydraulics and
geotechnical standpoint. I suspect that even if EPA had not done earthwork
near the GKM entrance, the plug of loose fill at the GKM entrance may
have eventually failed by internal erosion (piping) in a manner similar to
what occurred on 5 August 2015. As heads inside the mine rose, the
hydraulic gradient across the plug increased. Excavating activities also
increased the gradient across the plug, but it is quite possible that even
without that activity, the increasing gradient could have eventually
triggered a piping failure and a sudden release of water stored behind the
plug.
87) Much of the geochemistry followed well-accepted principles but there
were some exceptions. I have addressed these in my comments above.
EPA Response
86) The EPA/ORD team found
the background on the subsurface
hydrology of these mines compiled
by the US Bureau of Reclamation
to be very informative as well.
87) No response required.
50
-------
Question 10
Were precipitation and mineral saturation analyses of the acid mine drainage appropriately
applied for interpreting metals fate in the river system? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Kirk Nordstrom
88) For the most part, yes. First, the application of dissolved iron oxidation rates
was helpful to point out the enormous increase with pH. However, the fact that
microbes can speed up the rate enormously at low pH was not mentioned. This
should be mentioned along with the caveat that microbes would not have enough
time to develop sufficient colonies in the short time of the release to affect much
oxidation. There is often a 1-2 week lag time necessary before microbial colonies
are of sufficient concentration to show detectable changes in the ferrous iron
concentration.
89) Second, the saturation indices for calcite and dolomite were very pertinent and
appropriate. This is especially important in pointing out the neutralizing capacity of
the Upper Animas River. It would be really useful for the final report to do a
simple mixing with reaction calculations with PFtREEQC to simulate the effect of
mixing the estimated plume (at or near the peak GKM release) at the mouth of
Cement Creek and the Animas River to show the strength of the Animas in
neutralizing the plume. The geochemical analysis has made a start down this path
but a little more work should be done to complete this effort. I see it as a very
important part of the overall characterization.
90) The saturation indices for amorphous gibbsite was an appropriate figure, but it
is a little disturbing that the saturation state shows considerable over saturation for
many of the data points. We have not seen quite as much of this amount of
oversaturation. Because it is not reasonable to have such supersaturation relative to
freshly precipitating Al hydroxides, it should be assumed that some particles were
not fully filtered out. As pointed out for the large supersaturation for iron
hydroxides, considerable Fe particles get through the filter and indeed, may have
been formed during storage of samples if they were not filtered and acidified right
away.
88) We mention microbial
oxidation rate when discussing
Cement Creek and will make that
clear in the final report.
89) We agree with the suggestion
to use geochemical computer
coding to model reactions at
selected locations. Results of this
modeling will be included in the
final report.
90) The EPA/ORD team also
noted the spread on the gibbsite
saturation indices, including an
indication of supersaturation.
There are several possible
contributing factors discussed by
the group during the review,
ranging from filtering
inefficiency to estimation
techniques. This possibilities
will be further explored and
discussed in the final report.
51
-------
Question 11
Was the neutralization of acid mine drainage and subsequent fate of dissolved and
colloidal/particulate metals appropriately interpreted? Why or why not?
Reviewer Name
Reviewer Comment
EPA Response
Kirk Nordstrom
91) Yes, a very good start on the neutralization and fate of colloids
was done. As mentioned above, a PHREEQC simulation of mixing
with reaction to compare with the more qualitative description would
wrap this part up nicely.
92) The sorption calculations are considerably speculative, at least in
the way they were described. The ART should use Dzombak and
Morel's (1990) book on sorption to apply modeling because it is the
only place where a self-consistent set of data is available. I am still
not sure that scientifically justifiable results can come out of this but
at least this would be a starting point. Also, it should be noted that
Webster et al. (1998) EST 32, 1361-1368 found that the sorption of
acid mine drainage precipitates and schwertmannite were different
than ferrihydrite, shifting the sorption edge.
91) We will use geochemical computer coding
to model reactions at selected locations. Results
of this modeling will be included in the report.
92) Modeling of trace metal sorption on ferric
hydroxides was supplied only to offer
qualitative support of the conceptual model that
dissolved trace metals would be scavenged
from solution by the combined phenomena of
electrostatic and chemi-sorption on the
incipient ferric and aluminum hydroxides. We
will review Dzombak and Morel's book for
insight that may offer improvements to the
analyses and we will consider adding Webster
et al. to our discussion of sorption onto mineral
surfaces in our analyses.
52
-------
Part 4: Water Quality Analysis Simulation (WASP) Modeling
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
93) The methods and results for the WASP modeling
were unclear and it appears that a number of common and
well-accepted modeling practices were not used, with no
clear or acceptable explanation of why. Although I
understand that there is a lack of data in some areas and
that some time constraints contributed to the approach
used, I feel that the problem is important enough that the
best modeling approach possible should be used to derive
the most accurate and useful results possible.
Primary comments and issues are itemized below:
94) Although the primary goal of the modeling
was presented, this goal is very general and vague,
which leads to a great deal of generality and
uncertainty in the model results.
95) The segmentation and structure of the model
for surface water and sediment segments were
briefly discussed, but these should be clearly
presented in a map or schematic.
96) Although the modeled discharge was
calibrated to the flow measured at the USGS
gages, there was no attempt to include or model
major tributary inputs to, or irrigation or other
93) Through the peer review and associated comments, we
have updated the WASP modeling effort. These include
partitioning of individual metals in dissolved or particulate
form using WASP7 TOXI as well tight calibration of
model particle deposition to observations. We will fully
describe the implementation and calibration of WASP
along with calibration statistics in the final report. This
work will be further peer-reviewed before being finalized.
94) We will ensure that modeling goals are specific in the
final report.
95) A map for the model domain will be incorporated into
the report.
96) WASP modeling utilized USGS river gage data to
parameterize model hydrodynamics. Hydraulic geometry
relationships determining the depth and width exponents
of power equations were determined from the gage
available cross-section data (Leopold and Maddock,
1953). We calibrated the Manning's roughness
coefficients so that velocities matched observed flow
results.
We assume that irrigation influences were negligible
during plume movement since use had been curtailed,
53
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
takes from, the main stem of the rivers. These
should have been included. It is also not clear
whether and why both Geometry equations and
Manning's Equation using roughness and slope
were used to estimate flow parameters such as
velocities.
97) There appears to have been no attempt to
calibrate or validate the modeled concentrations or
loads. This is standard practice and there appears
to have been adequate data collected during and
immediately after the release to at least calibrate
the model, so this should have bene done.
98) It is not clear why the TOXI module of
WASP, for some reactive transport of toxicants
including metals, was not used. This module
includes Kd for partitioning between dissolved
and particulate forms, first order decay, and a
diffusion coefficient. Why were only total metals
modeled, whereas both dissolved and total (or
particulate) could have (and probably should have)
been simulated. Although WASP cannot model
equilibrium precipitation-dissolution reactions
based on pH and other parameters like some other
models (META4 and OTEQ), it can incorporate
Kd values and diffusion coefficients to simulate
adsorption and partitioning between the dissolved
which was clearly evident in the hydrographs for the
period. Flow volumes were adjusted at gages. Given the
abundance of gages, this was considered sufficient for the
scale needed to represent plume movement. We will
include discussion of factors affecting flow estimates in
the final report.
97) We have revised the calibration of WASP settling and
resuspension of total particulates based on empirical
estimates of metals loads during the plume using
observations at selected sites with multiple sampling. We
use the constant settling velocity based on Stokes' law. To
determine the settling particle size, we back-calculate the
particle sizes that would result in these settling velocities
that produce the loss of mass determined from the
empirical modeling of mass at sites based on observed
data. The methods used to estimate the loads from
observations as well as how they were used to calibrate
WASP will be fully described in the final report. This
method of calibration suggests silt-size deposition, which
appears reasonable. WASP was not calibrated to
individual samples of metal concentrations sampled in
what was thought to be the plume, given how limited this
data actually was and lack of clarity of exactly when each
sample was collected relative to the passage of the plume.
98) The WASP7 TOXI module was applied and described
in the peer review, which we make clear in the final report.
54
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
and solid or particulate phases with suspended
sediment in the water column and with bed
sediments. It appears that this should have been
done.
99) Also, why was the WASP add-on, Metals
Transformation and Assessment (META4), not
used for the fate and transport modeling? This
module was developed by EPA and can handle
reactive transport in complex acid mine drainage-
metals systems with precipitation-dissolution
reactions incorporating pH and other important
parameters.
We have incorporated Kd which is segment and metal
specific. The Kd is a lumped parameter of dissolved metal
and paniculate metal to capture precipitation and sorption.
We simulate total paniculate metals and the individual
metals As, Cu, Pb, and Zn, simulated as dissolved and
paniculate. Diffusion between the water column and the
sediment pore water has been incorporated. WASP
application with these additional elements will be fully
described in the final report.
99) We reached out to Dr. James Martin, one of the WASP
architects, to ask about META4. META4 is based on
WASP4 and was used on mine sites in CO by Medine in
the 1990s. To use META4, requires user knowledge of
META4 and WASP4. We are currently using WASP7.
The routines available in WASP7, such as the new
hydrology routines that allow for kinematic wave
modeling for streams/rivers, or the interface, are not
available in earlier versions of WASP. In addition,
META4 was never officially released and it does not have
the level of testing that WASP7 has undergone. An update
to META4 was done in 2012, but it was never been fully
implemented into WASP, and is not currently available or
usable. Dr. Martin recommended that META4 not be used
until it is finally incorporated into the updated WASP, and
then only if there is good water chemistry data. We do not
use META4 for this project.
55
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
100) It is not clear why Scenario 1 was used.
Although this scenario could provide a very
conservative estimate and upper bound on
transport estimates, this scenario may not be
conservative if scour of metals in sediment is
important, and it is too simplistic and not realistic.
101) It is not clear what initial and boundary
conditions were used for metals and sediment
(Total Suspended Solids) concentrations and
loads. For concentrations it appears that post-
plume metals concentrations were used, whereas
previous pre-release data are available and perhaps
should have been used.
102) Although sediment/particulate settling and
resuspension was included in Scenario 2, it is not
clear why there was no attempt to use any existing
information on surface (bed) sediment metals
concentrations for model input or calibration.
Some of these data may have been available from
previous studies to use as model input, and from
data collection afterwards for calibration.
100) Scenario 1 has been removed. We now simulate a
case where we only look at the total particulate load so we
can determine the movement of the plume independent of
other factors.
101) We set initial concentrations to 0 throughout the
system so that concentrations rose and fell solely due to
the passage of the release plume. This eliminates
background concentrations of dissolved calcium,
potassium, magnesium and sodium that are the primary
constituents in the water at this time of year. The upstream
boundary condition for total particulate metals in the San
Juan River is upstream of the confluence with the Animas
River and is based on observations in the lower Animas in
the San Juan immediately downstream from the
confluence.
Any estimates of background concentrations used in the
modeling are based on a few pre-plume samples or post-
plume data collected well after the plume passed. We have
evaluated post plume water quality trends and identified
that adjustments to water chemistry have occurred.
However, they are very small relative to the plume and
should not introduce significant error to plume modeling.
102) See Answer 97. We have revised the calibration
method by calibrating to computed metals mass based on
observed metals data.
56
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
103) In addition, there was no evaluation of
sediment particle sizes from any previous data for
modeling of settling and resuspension based on
velocity and shear stress. It is not clear why the
larger shear stresses in Scenarios 2b-2 and 2b-3
were selected and what kind of particle sizes these
may relate to.
104) For Scenario 3 for Long-term Effects, it is
not clear why Nov 2010 - Dec 2011 was selected,
what the magnitudes of the high flows modeled
were, and how representative they are of high
flows in these rivers.
103) The particle characteristics of the plume made up of
precipitates of iron and aluminum oxides and perhaps
gibbsite were probably unlike sediment that may have
been previously measured in most parts of this system.
Furthermore, the geochemical reactions during plume
movement would likely have caused changing particle
sizes as the plume traveled. We are not aware of any
particle data collected as the plume passed.
Because of the lack of data to support the erosion/settling
equations, of the GKM model, we have changed our
model formulation and are now using the dynamic erosion
equations to improve the settling functionality velocities.
We use the constant settling velocity based on Stokes' law.
To determine the settling particle size, we back-calculate
the particle sizes that would result in these settling
velocities that produce the loss of mass determined from
the empirical modeling of mass at sites based on observed
data. Estimated particle size (-silt) compares well with
what would be expected from precipitates in the plume.
The methods and calibration will be fully described in the
final report.
104) The Nov 2010 - Dec 2011 was chosen because it was
the most recent year with a complete hydrological record.
Coincidentally, it appears that spring snowmelt that year
may be very close to that occurring in 2016. The final
57
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
105) Using the model to estimate plume movement is
good, but using Scenario 1 (due to reasons stated above)
and qualitatively matching modeled to observed
concentrations visually are very questionable. Why were
no standard objective functions (such as Nash Sutcliffe or
correlation coefficients) used? This is standard practice.
report will provide the rationale for selection of the flow
record period used to model and analysis of its magnitude.
105) Scenario 1 has been eliminated. The final report will
show the relationship between observed data, empirical
reconstruction at selected locations and the WASP
simulations. Nash-Sutcliffe is a standard metric for
hydrological modeling, but not standard for water quality.
In hydrological models, you have a rich data set of
observed and simulated, while in water quality, you often
have limited data. We have included in our updated work,
the use of the correlation coefficient, R-squared.
Charles Fitts
106) Overall, the WASP modeling seemed to show that
the program could be made to simulate migration and
dispersion of the plume in the river that is fairly consistent
with observations of the plume's passage. It appeared to
be useful for simulating the approximate dilution and
dispersion of contaminants, but I felt less comfortable
with the analysis of deposition/resuspension since
required erosion/deposition parameters were far outside of
published ranges.
107) It was not clear to me how the equations for velocity,
depth, width (slide 9) and Manning's equation (slide 12)
were applied. Our 25 Feb phone call helped clear this up
for me, but that section could use clearer explanations
about how Q and V as a function of (t, river distance)
were calculated. If the velocity equation on slide 9 was
106) Following the peer review discussions, we have
changed our approach for settling/resuspension. See (103)
107) The power relations and their predicted exponents
from regressions were used to relate depth and width of
the channel geometry to flow. Flow and velocity were
calculated by solving the continuity and momentum
equations, incorporating Manning's equation. The figures
and equations will be improved in the final report.
58
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
not used, as Chris mentioned in our phone call, that
equation should be eliminated. Also, there are corrections
as noted in our discussion: the constants "a" and "b" in
slide 9 equations are duplicated but should not have been,
and the graph on that page needs axis labels.
108) It appears that Q was assumed constant from one
gage down to the next gage, where the Q abruptly jumps
up or down. It seems to me that it would be better to
assume a more gradual transition of Q from one gage to
the next, because the abrupt jumps in Q ripple through the
calculations to cause abrupt jumps in concentrations. If
you know where larger tributaries join, you could
improve the assumed distribution of Q between gages
using that knowledge. I understand from our phone call
that WASP is limited to 50 such discharge changes, which
is many more than were used in this simulation. So more,
smaller jumps in Q could be incorporated into the model
to give smoother, less distracting results.
109) Unless I am mistaken, the comparison shown on
slide 11 means little, since the model input was
constrained to match Q at gages, and all this slide shows
is that the constraint worked as expected for a spring
hydrograph record.
110) In slide 12 the discrepancies are not large, but they
are systematic - modeled velocities are consistently high.
108) To improve the incorporation of Q into the model, we
increased the number of inflows (boundaries) to refine
spatial representation of tributary inflow. We distributed
the change in Q by adding 3 more locations between gages
in the Animas River, and added 1 more location on the San
Juan.
109) This slide was presented to demonstrate exactly what
the reviewer points out~that our structuring of the flow
matched observations as expected.
110) We have further adjusted Manning's roughness
coefficients to produce both positive and negative
residuals. We will discuss how assigned Manning's
roughness are high due to the averaging nature of a WASP
segment, which does not capture the roughness effects
associated with reach scale channel roughness including
curving and braiding of the natural system.
59
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
The deviation should be explained, and the need for
unusually high Manning coefficients should also be
discussed and rationalized.
Ill) Note these typos in slides 15 and 16: the mass flux
between the segments n and n+1 should be "QnCn".
Numerical dispersion is noted in slide 15. If it is possible
to quantify that and compare it to simulated dispersion,
that would be helpful.
112) Slide 23 shows two graphs that appear to show the
same data. If there is a need for two graphs, explain what
they show and how they differ. If there are time-series
observed concentrations, they should be shown on these
graphs for comparison.
113) In slide 41, it would help to color code the sample
dots to indicate how close the measurement was to the
passage of the peak plume (e.g. red for within 1 hour of
peak, orange for 1-2 hours before or after peak, and so
on).
114) Slide 42 should show lines of relevant criteria other
than recreation, such as for aquatic life and drinking.
Ill) Typos will be corrected in the final report.
We will also include discussion of dispersion, both real
and numerical. Numerical dispersion causes the plume to
have an appearance like a normal distribution, when it may
be skewed and peaked.
112) Figures and graphs will be clarified in the final
report.
113) We will show the distribution of concentrations at
multiple sites during plume travel using color coding or
other means in the final report.
114) Plume concentration estimates will be compared to
water quality criteria in the final report.
60
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Henk Haitjema
115) For the most part it seems that proper use has been
made of available data and, when necessary, data from the
literature. However, this was not always fully explained
during the presentation. Specifically, the formulas
presented on slide 9, used to calculate the average water
velocity, stream depth, and stream width for a particular
model segment of the stream were not fully documented
(and contained some erroneous coefficients). A more
complete description of exactly what was measured where
and how the regression analysis was applied to arrive at
the coefficients "a" through "f" must be provided in the
final report.
116) In follow up discussions it appeared that WASP did
not use the first formula on slide 9 - the formula for
velocity. These velocities were obtained using Manning's
equation and calibration using observed velocities at
USGS gauges, see slide 12. This is of course confusing.
The velocity calculation on slide 9 is best removed. It
should also be explained why the calibration on page 12
left all observed velocities below the modeled velocities.
115) We have corrected equations and will fully explain
the data and method of analysis in the final report.
116) Velocities are determined with the Manning equation.
Reference to the velocity estimates based on hydraulic
geometry will be eliminated.
Kirk Nordstrom
117) It should probably be mentioned that there are other
transport codes for this situation (e.g. OTEQ and
PHREEQC) and some justification should be given why
the team used WASP instead of something else.
Especially in light of the fact that OTEQ has been used on
mountain streams containing acid mine drainage for about
20 years and PHREEQC has been used longer than that
117) We will discuss in our presentation of WASP and
model selection the availability of other models, such as
OTEQ. We used WASP because we have expertise in this
model and felt that it could represent the travel
characteristics associated with basin hydrology (dilution),
local hydraulics and deposition of a plume of AMD in a
river system 600 km in length. We recognize that
61
-------
Question 12
Did the WASP modeling appropriately apply modeling parameters to
estimate the movement of plume water? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
for geochemical modeling of acid mine drainage
chemistry.
118) I was glad to see that the team did not try to combine
transport with reaction because there is not sufficient data
to constrain such modeling.
important geochemical transformations of plume dissolved
metals would take place, especially in the first hours of
plume travel once the plume passed into the Animas River.
The suggested META4 module would potentially have
assisted with this but it is not currently publicly available
or incorporated into WASP7. Such a modeling effort at
this geographic scale would also have been hampered by a
lack of geochemically-relevant data everywhere in the
system of interest. Instead, the EPA/ORD team relies on
geochemical analyses conducted partially through
independent geochemical equilibria models and supported
by what data was available. We believe for purposes of
characterizing transport, fate, and exposure that the
resolution achieved with this approach at this scale of
analysis meets project objectives.
118) The EPA/ORD team agrees that to combine reaction
with transport is ill-advised given the limited data.
62
-------
Question 13
Did the application of assumptions and values in WASP modeling appropriately
address particle transport and deposition of the acid mine drainage constituents? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
119) This comment addresses both question 13 and 14
below, since they are related. It is not entirely clear
how the application of assumptions and values in
WASP modeling addressed particle transport and
deposition of the acid mine drainage constituents. As
stated above, although sediment/particulate settling
and resuspension was included in Scenario 2, it is not
clear why there was no attempt to use any existing
information on surface (bed) sediment metals
concentrations for model input or calibration. Some of
these data may have been available from previous
studies to use as model input, and from data collection
afterwards for calibration. In addition, there was no
evaluation of sediment particle sizes from any
previous data for modeling of settling and
resuspension based on velocity and shear stress. It is
not clear why the larger shear stresses in Scenarios
2b-2 and 2b-3 were selected and what kind of particle
sizes these may relate to.
120) It is not clear exactly what the settling results
(slides 25-28) of the WASP presentation are showing
or how they were computed. There appears to be total
metals concentrations in the water column, and a
certain fraction of this mass settles out into the
sediment, creating total metals concentrations in the
surface sediments, a certain faction of which can be
re-suspended under high flows. But these are
119) Given an improved understanding of the system,
particularly the calculated loads and decrease in load traveling
downstream,
As also described in (103), we have based the model
formulation on a constant settling velocity rather than the from
dynamic erosion algorithm presented to the review panel.
Settling velocity is determined from using the empirically
quantified plume mass based on observed data at selected
locations. The calculated settling rates are calculated using the
particle sizes that would have these settling rates. The settling
velocity is applied to the reaches between calibrated locations.
The EPA/ORD team feels that the movement of this plume
through the system was a unique event whose conditions
would not be well represented in existing data. Methods for
model parameterization has generally described above will be
fully explained and compared to existing data to the extent
possible in the final report.
120) WASP outputs results representing sediment
concentrations in non-standard units. In the final report we
will report WASP output in standard units of mg/kg.
Assuming low flow and suspended sediment on the day of the
GKM plume on August 5, the initial concentrations of
sediment were assumed to be 0 for the analysis.
63
-------
Question 13
Did the application of assumptions and values in WASP modeling appropriately
address particle transport and deposition of the acid mine drainage constituents? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
presented in units of mg/L instead of mg/kg, which is
typical for metals in sediment. Typically you would
have particulate metals in the water column which can
settle out. In addition, as stated above there appears to
be no initial metals concentrations or mass in the
surface sediments as an initial condition, although in
some locations there should be some of this type of
data from previous studies. It is also not clear if there
was any initial suspended sediment loads or
concentrations in the water column as part of initial
conditions.
Charles Fitts
121) The estimation of the GKM release load (WASP
slide 13) is discussed in question 6 above.
The WASP model simulated only the metals loads in
the discharge of Cement Creek where it joins the
Animas. It assumed zero metals loads in the Animas
above Cement Creek and in the San Juan above the
Animas. It would make for more meaningful
comparisons with observations (e.g. slide 22) if
estimated loads from the San Juan and upper Animas
were added to the model. Adding these inputs may
also impact the simulations of settling and
resuspension. I imagine that reasonable estimates of
the upper Animas and San Juan loads could be made
from longer-term monitoring data.
121) During the summer low flow when the GKM release
occurred, metals constituents in the Animas River are
dominated by dissolved calcium, potassium, magnesium and
sodium with low or non-detectable levels of the key metals of
interest that were significantly elevated in the GKM plume.
We felt that inclusion of "background" metals would not
improve modeling results. However, major cations introduced
with the mine effluent were treated as plume load while
background concentrations of major cations were ignored. As
noted by the reviewer, WASP models the GKM metals plume
as if it were moving through distilled water.
WASP loads were adjusted as the plume entered into the San
Juan to represent the large sediment load from upstream that
joined the GKM plume at the junction of the Animas and San
Juan Rivers. In the updated simulations, Kd is applied as a
lump parameter for precipitation reactions and sorption to the
64
-------
Question 13
Did the application of assumptions and values in WASP modeling appropriately
address particle transport and deposition of the acid mine drainage constituents? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
122) In slide 2 (and 28), the 3rd item says ".. .upon
entering Cement Creek", but I think that should be
".. .where Cement Creek enters the Animas River."
The last item on these slides should start with
"Simulations indicated that high flow periods..." to
clarify that this is a simulation result, not a measured
result.
123) The simulations of sediment deposition and
subsequent re-suspension indicated that a likely
stretch of river to receive such deposition was from
about 65km to about 95 km (WASP slides 25-27).
However, the 2b-2 simulation used erosion/deposition
critical shear stress thresholds that were far outside
expected ranges, thus the results should be viewed
only as qualitative. There should be a discussion
explaining possible factors that required such large
thresholds, and the degree to which these factors
render the results useful or not.
124) One problem with the WASP modeling of
erosion/deposition is that it must treat long stretches
of river as homogeneous with respect to velocity,
which is far different than the actual riffle and pool
nature of the Animas River where it occupies an
alluvial plain. This point is also discussed in question
7. In think plume sediment settled out in the calmer
pools to a greater degree than what the WASP model
particulate metals. We estimated the contribution of
particulate metals coming in from the San Juan above the
Animas by comparing the total particulates in the San Juan
downstream of the confluence with the Animas and the total
particulates in the Animas before the San Juan.
122) These suggested phrasings were updated in the final
report.
123) The treatment of resuspension/deposition has been
improved. See (103) and (119).
124) The EPA/ORD team agrees that an average reach
velocity used by WASP does not represent the local scale
variation in velocity that exists and would likely contribute to
deposition of the plume particulates as it travelled.
Undoubtedly, there was local settling of colloidal/particulate
precipitates and particles in the low velocity zones that occur
near the bed and banks, and locally within the riffle/pool
morphology, as verified in photographs. WASP is not readily
applied to finer scale channel morphology and streambed
features that contribute to heterogeneities in roughness and
velocity. We agree that it is important to highlight this in the
final report.
65
-------
Question 13
Did the application of assumptions and values in WASP modeling appropriately
address particle transport and deposition of the acid mine drainage constituents? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
would predict for homogeneous segments. This may,
in part, explain why atypical erosion/deposition
parameters were required to make the simulation
match observations.
Henk Haitjema
125) It offered a first approximation to these
processes. I appreciate the use of a simple modeling
approach as conducted here in view of the limited data
availability and the limited study objectives as
described by the EPA team leader and the modeler. In
particular, I applaud that only the most fundamental
processes have been included, while secondary
processes that are more difficult to parameterize have
been omitted. In this light I agree with the decision to
ignore physical dispersion in this modeling exercise.
However, some caution is needed to declare the
omission of (physical) dispersion conservatively by
declaring that the modeling results provide an "upper
bound." It does for the concentrations (assuming that
numerical dispersion in the model does not simply
replace the physical dispersion in the river or even
exceeds it), but it is not conservative in predicting
early arrival, for instance. The fact that numerical
dispersion has not been quantified relative to physical
dispersion is a weakness in this study.
125) The EPA/ORD team agrees that the simple modeling
approach is the best approach given our objectives and limited
data sources. We further agree that numerical dispersion does
not necessarily result in an "upper bound". There is
insufficient data to adequately quantify the actual physical
dispersion that occurred as the plume travelled or compare it
to the numerical dispersion produced by the model. We will
clarify the effects of dispersion and include a discussion of
uncertainties associated with the modelling in the final report.
Kirk Nordstrom
126) As a more general comment - it would seem to
me that putting the metals concentrations (dissolved
and total) and loads in the perspective of the range of
all data for low flow conditions (or similar flow and
126) We have increased use of historical data to help put our
simulation results into context and to evaluate post-event
water quality after the plume passed. While the plume
concentrations were very high, easily observed and quantified,
66
-------
Question 13
Did the application of assumptions and values in WASP modeling appropriately
address particle transport and deposition of the acid mine drainage constituents? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
time of year as the GKM release) would help put the
plume release in better perspective. This is where
historical data could help considerably. A max and a
min from historical data could show some kind of
envelope around or near the plume results.
127) Also, it would be better to show the individual
metals, especially Cu, Zn, Pb, and As when
comparing the peak concentration with river distance
and conservative (no settling) scenarios. It is also not
clear why some samples that look like they were
sampled nearly the same time had such different
concentrations. This graph needs a lot better
explanation.
at least in the Animas River, the historical comparisons are
especially important in trying to discern if the river system has
returned to pre-event levels.
127) In response to the reviewer's comments we have added
simulations for individual metals including Cu, Zn, Pb, and As
to our general treatment of summed metals. The observed
concentrations are presented to provide context for the
simulated results. For a given date, there may be multiple
samples at the same locations, collected at different times on
that date. The bulk of the plume mass moved through each site
rather quickly and samples collected even a few hours apart
often had very different concentrations. This is fully explored
elsewhere in the project analyses.
67
-------
Question 14
Did the WASP modeling appropriately investigate the remobilization of
metals during increased flow? Why or why not?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
128) Please see comment on question 13 above.
128) See response to comment (103) and (119)
Charles Fitts
129) Since the mechanisms for erosion/deposition in the model were
using critical shear stress thresholds that are well outside normal
ranges, these results must be viewed with skepticism. Like I said in
the previous comment, this makes quantifying concentrations very
uncertain and this analysis should probably be viewed as an
example of how downstream concentrations could respond during a
high flow period, not a prediction of how they will likely respond.
That distinction should be made clear in the text.
130) On slide 27, label that these are scenario 2b-2 results, and
correct the title of the lower graph so it says "Movement of Total
Metals..." The resuspension scenario needs to be outlined clearly.
What is the assumed event? Is it a hydrograph from a typical spring
runoff period, a shorter duration storm event, or something else?
129) We have changed how we handle the
mechanism of settling and resuspension. See
response to comment (103) and (119). We agree
and will present results as a hypothesis of what
could happen rather than what will happen. The
amount of resuspension and the nature of
suspended materials in future runoff events
remains uncertain.
130) The year 2010 - 2011 hydrology record was
used for the simulation. The final report will
discuss how this this year relates to other years.
HenkHaitjema
131)1 am not sufficiently familiar with the WASP model to
adequately evaluate this point. However, I do have an observation
on the reconstruction of the total metals concentration in the release
flow from the mine as presented on page 13. This calculation was
done outside of WASP. In principle, the mass balance calculation as
presented by the three formulas on page 13 is elegant due to its
simplicity. What became apparent during the discussion, however, is
that the assumption that the total metals concentration in the plume
as measured at 4 p.m. at the 14th St. Bridge, which is after the peak
of the plume passed, is the same as during the peak flow (peak of
the plume) at 12:45 may be problematic. Reviewers pointed out that
the higher turbulence in the peak flow more likely than not would
have caused much more materials in suspension and thus a (much)
131) The analysis presented to the panel stated
that the 4:00 pm (16:00 hr) sample was not
equivalent to the peak and a method was offered
to estimate (raise) the concentration to a more
representative value. The panel and EPA/ORD
team had considerable discussion on how to do
that. The final report will rectify any figures that
are unclear on this point. We have also corrected
an error in the algorithm for total concentration
and we now assure that both the WASP and
empirical modeling of the plume begin in Cement
Creek at the same location and that they use the
same peak value. We feel that our analysis
68
-------
Question 14
Did the WASP modeling appropriately investigate the remobilization of
metals during increased flow? Why or why not?
Reviewer Name
Reviewer Comment
EPA Response
higher total metals concentration than what was measured at 4 p.m. I
concur with that observation and emphasize that the assumption on
slide 13 (concentration at 4 p.m. same as in during peak flow) might
significantly underestimate the total metals load that entered the
Animas River! Thus the problem with this assumption is that it is
not conservative as to the study objective, assessing the potential
impact of the release.
132) However, the total metals load in the plume further down
gradient in the Animas River has been independently estimated from
plume size and concentrations there. Thus, the impact assessment of
the release downgradient in the Animas River is not dependent on
the estimate of the original total metals load in the plume while it
was still in Cement Creek. It is more likely than not that the missing
portion of the total metals load in Cement Creek (due to the
underestimation discussed above) ended up as sediments in the very
first kilometers of the Animas River. This is because the flow
velocities were quickly reduced as soon as the plume entered the
Animas River thus allowing settlement of the larger particulates that
might have made up the higher peak flow concentrations.
133) In summary, while I agree that the peak flow concentrations in
the plume in Cement Creek may have been underestimated, I do not
believe that this underestimation affected the down gradient impact
assessment, except perhaps for the sediment load in the first few
kilometers of the Animas River.
choices result in "conservative" estimates, but we
agree that cannot be known with certainty and
estimates could be too high or too low. Therefore
we will remove any text that suggests our
modeling choice is clearly conservative.
132) We agree that some portion of the metals and
sediments entrained in what was a moderate flood
in Cement Creek would have certainly deposited
within the first kilometers of the slower-moving
Animas River. As noted, the implication of
under-estimating the peak concentration in
Cement Creek is that there would have been more
actual sediment deposited in the Silverton reach
than modeled. The final report will discuss the
uncertainty in mass and how they affect estimates
of where deposition occurred.
133) We agree that any underestimation of the
peak concentrations in Cement Creek would result
in underestimation of deposited material between
A72 and Cement Creek, but that an error would
not have affected the estimated load still carried in
the river past Silverton (A72). We will fully
discuss the implications of this in the final report.
69
-------
Part 5: Groundwater Modeling
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
Henk Haitjema
134) In principle it is. While a very basic groundwater code (WhAEM)
was used, its limited capabilities are consistent with both the very
limited data available and the limited objectives of this study. In other
words, a more sophisticated model would require additional
assumptions and would, therefore, not have offered more insight. To
fully assess the sufficiency of the current analysis it is necessary to
consider its objectives. The ultimate question to be answered is (from
the presentation): "Could drinking water or irrigation wells drawing
from river alluvium become impacted from the chemicals associated
with the GKM release?"
This question may be broken up into three interrelated questions:
a) Which wells, if any, receive some of its water from the
river?
b) What are the travel times of water from the river to those
wells?
c) What is the dilution in the well of possible contaminants
received from the river?
Question (a) can be answered with capture zone analyses for the various
wells. Question (b) can be answered by use of forward particle tracking
starting at the river and ending in the well. Question (c) can be
answered by tracing particles backward in time from the well, using a
uniform distribution of particles around the well, and then comparing
the number of path lines that reach the river to those that do not.
134) We have expanded the modeling
approach and acquired additional data to
better inform calibration. The refined
approach creates a floodplain scale
("regional") model to set the boundary
conditions for a local model capable of full
3-dimensional flow characterization applied
at selected wells. The updated regional
analytic element GFLOW model is
calibrated to the observed static water levels
at nearly 300 wells in the nearfield and
baseflow estimated from USGS gages,
giving effective hydraulic conductivity for
the alluvium and surrounding rock, and an
effective regional recharge rate. The local
scale MODFLOW model with boundary
conditions informed by the regional model
are constructed for selected wells to better
explore the local-scale complexities. This
modeling combination is used to explore
localized effects of factors such as
heterogeneities in floodplain deposits (e.g.
paleochannels), the effect of irrigation
ditches in controlling floodplain water
levels, as well as the impact of transient
pumping and well interactions. In turn,
70
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
WhAEM is EPA's standard model for well capture zone delineation in
the context of wellhead protection and can address all three questions.
As such it is a logical choice for this analysis. However, it is necessary
to consider both the limitations of WhAEM and the limitations in field
data, and document how these might impact the outcome of the analysis
with the above research questions in mind. I will discuss these
limitations in arbitrary order below.
Dupuit-Forchheimer flow
WhAEM falls in the class of codes that solves "two-dimensional flow in
the horizontal plane," at least that is how these types of models are
routinely referred to. Regretfully, this is misleading terminology!
WhAEM is a Dupuit-Forhheimer model, which is a model in which
resistance to vertical flow is being ignored, thus not vertical flow itself.
While the underlying partial differential equation in WhAEM involves
only the horizontal coordinates (x and y), flow into the vertical direction
can and is being approximated using conservation of mass
considerations. Consequently, path lines in WhAEM are being traced in
three dimensions.
For a Dupuit-Forchheimer model to offer a good approximation to the
actual three-dimensional flow regime, its application must be limited to
groundwater flow systems in which the horizontal distances traveled by
groundwater are much larger than the vertical distances traveled. In
practice, this translates into groundwater flow systems in which the
distances L between boundary conditions (e.g. distance of the well from
the river) is larger than five times the aquifer thickness. This is for
isotropic aquifers. In case the aquifer is anisotropic, with a lower
insights from the local model are used
iteratively to inform the Animas River
floodplain-scale model.
The revised modeling approach using
GFLOW and MODFLOW will be fully
described in the final report. The report will
also contain sensitivity and uncertainty
analysis.
71
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
vertical hydraulic conductivity than the horizontal conductivity, the
following criterion may be used (Haitjema 2006):
L>5H
Where //is the aquifer thickness, kv is the vertical hydraulic
conductivity, and kh is the horizontal conductivity. Dr. Fitts suggested
using a kh to kv ration of 10 for the Animas alluvium, which seems
reasonable to me. The condition in the displayed formula above is not
meant for wells that are relatively close to the Animas River, and
unfortunately these are the wells of most interest (most likely to receive
river water).
What is the consequence of violating the Dupuit-Forchheimer criterion
for wells near the river? In reality the well - river interaction is
influenced by possible (bottom) resistance to flow between the river and
the aquifer as well as resistance to vertical flow inside the aquifer.
Neither is included in the model presented, although bottom resistance
could have been applied. By not including any of these resistances, the
flow potential for drawing water from the river that flows into the well
is overestimated. In other words, the model as constructed is
conservative with respect to the objectives of this study. To keep the
analysis conservative in nature I recommend not adding bottom
resistance to the line-sinks representing Animas River.
135) Still to be done: While the analysis as conducted and presented is
sound regarding this issue it must be fully documented, including
calculations for representative wells to show whether they satisfy the
Dupuit-Forchheimer criterion or not. This is currently missing from the
135) We have obtained better data on
aquifer thickness based on a gravimetric
survey of the mid Animas alluvium. As
suggested by the reviewers, we re-examined
the Dupuit-Forcheimer assumption for the
modeled wells using improved estimates for
aquifer thickness and the assumption of
anisotropy in hydraulic conductivity. For
one of the modeled wells, the DF
assumption was violated, confirming the
reviewer's assessment. The modeling
approach employed in the study has been
expanded to include models that can address
fully 3-D flow and improve regional water
table calibration (GFLOW). See (134).
72
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
analysis! A sketch of the expected three-dimensional flow patterns and
those modeled in WhAEM, as suggested by Dr. Fitts, would be useful
for full disclosure of this issue. Remember, that the path lines in
WhAEM are also 3D, but approximate and thus somewhat different
from the "real" 3D path lines in cases where Dupuit-Forchheimer does
not apply. What the sketch would not show is the fact that vertical
resistance to flow is ignored in the Dupuit-Forchheimer model, which
must be made clear in the figure caption.
136) Single homogeneous aquifer with horizontal base
WhAEM represents the alluvium near the Animas River as a single
homogenous aquifer, which means that it lumps the various depositional
layers in the alluvium into a single homogenous layer. Furthermore, it
assumes a horizontal aquifer base below which no flow is considered.
The question is how these simplifications affect the modeling results.
Specifically, what effect does this simplification have on the potential
well - river interaction? This was not discussed in the presentation, but I
will address this below.
The actual aquifer base is unknown, but at or below the depth of the
wells in the alluvium. In the absence of data it has been assumed in the
current analysis that the aquifer base occurs at the bottom of the well
under consideration. I agree with this choice! This will generally lead to
an underestimation of the aquifer thickness, but does not affect the flow
regime as much since the transmissivity in the model does not depend
on this assumption because it has been based on a pump test. Assuming
for a moment that the transmissivity is accurate (or reasonable) an
underestimation of the aquifer thickness will result in an overestimation
13 6) The EPA/ORD team agrees with
comments regarding the significance of the
assumption of a homogeneous aquifer with
horizontal base on the groundwater capture
zones and particle tracking. Estimates of the
base and thickness of the floodplain aquifer
has been improved based on an Animas
valley gravity survey (Hasbrouk
Geophysics, 2003). The actual aquifer base
and elevation for the local scale groundwater
models of the community wells remains an
estimate that will be explored through
sensitivity analysis. These data and the
sensitivity analysis will be described in the
final report.
73
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
of the hydraulic conductivity, since the product of the two is the
(known) transmissivity. So while the discharge rates in the aquifer,
including the flow component from the river if present, are not affected
(question a), the specific discharges and associated average
groundwater flow velocities are. An underestimation of the aquifer
thickness will result in an underestimation of the groundwater travel
times (question b). This is conservative in view of the model objective
since actual early arrival of contaminants may be later than predicted by
the model.
The actual aquifer heterogeneity offers the potential for preferential
pathways from the river to the well. The WhAEM model assumes a
homogeneous aquifer that lacks preferential flow. Consequently, the
assumption of homogeneity is not conservative in view of the model
objectives. Preferential pathways would shorten the travel times from
the river to the well (question b). While a multi-layer model may be
able to capture this effect to some degree, data on aquifer stratification
near the wells or between the wells and the river are absent.
137) Still to be done: The above discussion must be integrated into the
description of the modeling analysis to fully disclose the impact of the
simplifications and assumptions. It should be pointed out that predicted
early arrival times in the wells of chemicals released from the river may
not preclude that some (small) portion of the chemicals arrive even
earlier due to preferential flow. This is true in spite of the fact that the
actual aquifer thickness may be larger than assumed and thus result in
slower groundwater velocities, hence later early arrival than predicted
137) The final report will include a
discussion of the implications of preferential
flow pathways on breakthrough curves. A
preferential flow pathway associated with a
buried stream channel was introduced to the
model of a community well to demonstrate
the point.
74
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
by the model. Preferential flow may well outweigh the effect of the
aquifer thickness on the groundwater velocities.
138) Steady state flow
WhAEM models steady state flow, ignoring water that may go into
storage or is released from storage due to temporal changes in the water
table (unconfined flow) or head (confined flow). For the purpose of
capture zone delineation (in the context of wellhead protection), a
steady state model is considered adequate. In fact, producing capture
zones that change over time seems impractical for the purpose of
defining wellhead protection areas. However, replacing the actual
transient flow system by a steady state one raises the question what the
steady state model actually represents. Haitjema (1995, 2006), using a
study by Townley (1995), presents a dimensionless response time T:
SL2
T =
4TP
where S [-] is the aquifer storage coefficient, L [m] the distance between
head specified boundaries, r[m2/day] the aquifer transmissivity
(product of aquifer thickness and hydraulic conductivity), and P [days]
the period of a periodic forcing function. This formula differs slightly
from the one presented on slide 12 due to a different definition of the
distance L. When considering seasonal variations in flow in the alluvial
aquifer, the definition of L on slide 12 is more convenient where it is the
distance between the river and the valley boundary (rock outcrop).
Haitjema (2006) offers the following rules of thumb:
T < 0.1 treat transient flow in the aquifer as successive steady state.
13 8) The EPA/ORD agrees that the rule-of-
thumb analysis using the dimensionless
response time parameter, T, provides
guidance on how best to use the steady state
model with the available data, or when a
fully transient model is more appropriate.
The formula for dimensionless response
time, T, as well as an error in pump test data
for storativity were corrected. This brought T
to a range of 0.29 to 0.36, which is more in
line with what would be expected in
unconfined floodplain deposits.
The updated analysis shows that T is most
sensitive to the assumed periodicity. If P is
assumed to be 365 days, which represents
the annual snowmelt forcing, then T is
sometimes < 0.1 and other times between
0.1 and 1. When T < 0.1 this suggests
putting in actual pumping rates as successive
steady states. When T is between 0.1 and 1,
this suggests putting in both actual pumping
rates and averaged pumping rates to bound
the response.
Because the community well falls in the
mid-range, where transient flow may not be
well represented by a steady state model, the
75
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
0.1 < T < 1 transient flow cannot be meaningfully represented by a
steady state model.
T > 1 represent transient flow by a steady state model using average
boundary conditions.
These guidelines are approximate in that values just below 0.1 or just
above 1 are to be considered transitional from the aquifer responding
relatively fast or slow to transient forcing, respectively.
The analysis offered on slide 12 is incomplete and partially incorrect!
First of all, successive steady state simulations requires T to be smaller
than 0.1, not 1 as stated on the slide. Incidentally, I missed this issue
during the presentation. Secondly, the parameters used in the equation
on slide 12 must be reassessed. Specifically, the storage coefficient S is
too small for an unconfined aquifer as present in the alluvium
surrounding the Animas River. Instead of ^-values between 0.003 and
0.006 as shown on slide 12,1 expect the ^-value to be more in the order
of 0.1. This will increase the dimensionless response time T by almost
two orders of magnitude!
Note: The reason for the measured low ^-value is unclear, although Dr.
Fitts suggested that it may be an artifact of an imperfect pump test and I
concur.
Secondly, while the periodicity of P=365 days is appropriate to assess
the response of the flow system to seasonal variations in recharge (in
this case inflow into the aquifer near the rock outcrop) and seasonal
variations in river stages, it is not suitable to assess the response of the
flow system to short term variations in pumping and short term
final report will include an investigation of
the steady-state vs transient assumption on
capture zones for the community well that
has pump test data.
76
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
variations in river stage (e.g. storm surges). For that purpose a
periodicity P=\ day would be a better choice. This reduction in the
value of P would further increase the value of T indicating that the
aquifer responds rather slowly to storm events and pumping variations.
The current analysis does not distinguish between these long term
(seasonal) and short term effects (storm events and pumping variations).
Therefore, the current analysis mistakenly suggests that successive
steady state is always a good approximation of this transient flow
system, while in fact it most likely is not.
Still to be done: The analysis on slide 12 has to be redone with
appropriately chosen parameters and for different combinations of
parameters that apply to different (well) locations and include
bracketing values for those parameters that are not fully known from
field data. None are, of course! It is likely that most of these new
calculations of T will result in values larger than 1 or at least larger than
O.I.I recommend that for values larger than 1 average pumping
conditions and average river stages are used in delineating well capture
zones. In addition, I suggest repeating the analysis for the actual river
stage during the passing by of the plume. I suggest that for values of T
smaller than 1, but larger than 0.1, both average pumping rates and
actual pumping rates are used (pumping rates when the wells were
turned on). Finally, I suggest producing capture zones for bracketing
values of uncertain parameters. The resulting suite of capture zones
would account for (1) representing transient flow as steady state and (2)
for data uncertainty.
77
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
139) Groundwater levels and calibration
In almost all cases a groundwater flow model is being calibrated using
observed potentiometric heads (confined flow) or water table elevations
(unconfmed flow). Ideally, base flows in streams are also included as
calibration targets. Calibration leads to the determination of most likely
hydrogeological parameters such as hydraulic conductivities, aquifer
recharge due to precipitation, and perhaps stream bottom resistances. In
the current study area (or areas) almost no water level data were shown
to be available.
I wonder, however, if the many domestic wells in the alluvial aquifer
may be on record with the state (well logs). If so, many of them might
have static water levels that can be used as calibration targets. Similarly,
the high capacity wells (irrigation wells and public water supply wells)
may have logs that include static water levels as indeed are shown on
slide 11. In the absence of domestic well static water levels, the static
water levels in high capacity wells could be used as calibration targets
by excluding them one-by-one from the model. This would mean
conducting several calibration runs each with one of the high capacity
wells replaced by a "test point" (calibration target).
Currently, hydraulic gradients toward the Animas River are generated in
the model by defining head specified boundaries away from the river.
The water released by these head-specified boundaries presumably
comes from the surrounding mountains. A common approach in
modeling flow in alluvial valleys is to apply so-called "mountain range
recharge" along the valley boundaries at the bottom of the surrounding
mountains. In WhAEM this would be done using discharge-specified
line-sinks along the base of the mountains or boundary of the alluvium.
This, of course, is only possible if there are reasonable estimates
available for the mountain range recharge rate. However, the measured
baseflow increase along the Animas River could offer some insight into
78
139) The EPA/ORD team agrees with the
comments on the issue of model calibration.
We have expanded modeling to include
regional calibration using GFLOW (see 134)
using observed static water elevations
reported by almost 300 wells. The GFLOW
model and its conjunctive groundwater-
surface water solution technique allows the
additional calibration to the estimated
baseflow gain between the USGS gages on
the Animas River that bracket the study
area.
The regional model set the boundary
conditions for the local capture zone models
(MODFLOW) where the sensitivity analysis
was expanded to include porosity and local
hydraulic conductivity. We will discuss the
uncertainties in groundwater flow rates
towards the Animas, including accuracy of
elevation data, and we will seek data that
may improve estimates. A recent synoptic
well water level study with LiDAR-based
altimetry is available for the lower Animas
River floodplain study area that will help
overcome some of these limitations when
modeling this area.
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
this mountain range recharge rate. This baseflow increase has already
been considered in the current study as shown on slide 8.
Still to be done: The final report on the WhAEM model study should
address the uncertainty in groundwater flow rates toward the Animas
River and the lack of calibration targets. Possible data sources, as
mentioned above, should be discussed and used if data is indeed
available. The data uncertainty should be resolved through sensitivity
testing. Different possible groundwater flow rates toward the river
should be tested in the model and their effect on the capture zones
(question a), early arrival times of chemicals in the wells (question b),
and dilution of chemicals in the well (question c) be shown and
discussed. Note: The hydraulic gradient toward the river in combination
with the aquifer transmissivity defines the groundwater flow rate toward
the river. This groundwater flow rate is what really matters for the shape
and orientation of the capture zones, addressing questions (a) and (c).
The groundwater flow rates toward the river may be bounded by
considering base flow increases along the Animas River as already done
on slide 8. The actual hydraulic conductivity as well as the aquifer
porosity, however, affect the groundwater flow velocities and thus
travel times (question b). A sensitivity analysis on hydraulic
conductivity and porosity are thus also in order.
References
Haitjema, H.M. (1995) Analytic Element Modeling of 'Groundwater Flow. Academic
Press, San Diego.
Haitjema, H.M. (2006) The Role of Hand Calculations in Ground Water Flow
Modeling. Groundwater, Vol. 44, No. 6, pages 786 - 791.
79
-------
Question 15
Is the analysis as presented sufficient to evaluate the potential for impact of the acid mine release
from the GKM on pumping wells located in the floodplain aquifers downstream of the spill?
Reviewer Name
Reviewer Comment
EPA Response
Townley, L.R. (1995) The Response of Aquifers to Periodic Forcing. Advances in
Water Resources 18: 125 - 146
Charles Fitts
140) I think the analysis is sufficient to conclude that certain wells
farther from the river were not susceptible to drawing river water (wells
575m71km, 650m71km, lOOOmTOkm), and that certain wells close to
the river likely do draw in river water (well 35m66km and the 5 NM
wells). The analysis of well 75m71km was a closer call, and for that
well a more sophisticated analysis could shed better light on the extent
to which it draws in river water. The analysis conservatively estimates
whether wells have potential for drawing river water. By conservatively
I mean that the potential for drawing river water and the fraction of
water drawn from the river is probably overestimated by the WhAEM
models. This is discussed in more detail under question 16.
140) We agree with the assessment of the
likelihood of community wells drawing
source water from the Animas River. To
address the issues described here, as well in
Question 16, the groundwater modeling
approach has been expanded to include
models that can address local 3-diminesional
flow paths (see 134).
80
-------
Question 16
Were the assumptions informing the choice and construction of the groundwater flow model
appropriate for the intended use? Please explain.
Reviewer
Name
Reviewer Comment
EPA Response
Henk Haitjema
141) In general they were, but additional analyses and discussions are needed
as indicated in my answer to question 15.
141) See the EPA/ORD response to
Question 15 (134).
Charles Fitts
142) As stated above, the potential for communication between well and
river is probably overstated in the WhAEM model results. The main reason
is that the models are two-dimensional, not three dimensional. The models
neglect the resistance to vertical flow and only account for resistance to
horizontal flow. The models treat the river boundary condition as a fully-
penetrating vertical curtain of specified head extending from the surface to
the bottom of the aquifer. In reality, the river bed is a nearly horizontal
constant head boundary atop the aquifer, and the well screen elevations are
unknown but probably in the lower portion of the aquifer. For wells closer to
the river, vertical resistance is a significant part of the total resistance to flow
between the river bank and the well screen. Vertical resistance is even
greater in stratified sediments where the vertical hydraulic conductivity (Kv)
is much smaller than the horizontal hydraulic conductivity (Kh). For
example, consider well 35m66km which is 35 meters from the river
(horizontal) with a well screen that is about 25 meters below river. If Kh/Kv
= 10 (ratios of 5 to 50 are common), the vertical distance in an equivalent
isotropic medium would be about 80 m vertical distance (scale the vertical
axis by the square root of Kh/Kv to make an equivalent isotropic medium).
In this case, the vertical resistance between river and well screen would
likely be greater than the horizontal resistance. Neglecting the vertical
resistance in the WhAEM models overestimated the communication between
well and river, and underestimated the travel time for flow from river to well.
143) The models presented are very large scale and may be omitting some
important features closer to the wells. The alluvial plain has irrigation
ditches, old braids, and other surface water features that likely connect to the
142) We recognize the issue of
neglecting resistance to vertical flow in
the analytic element model (WhAEM).
We conclude based on the input of both
reviewers (Haitjema, question 15) that we
cannot assume 2-D flow as represented in
WhAEM to investigate communication of
wells near the river without independent
verification. We have expanded the
modeling approach to include a regional
scale calibration (GFLOW) and local 3-D
representation at selected wells
(MODFLOW). See (134).
We agree that the floodplain deposits are
likely stratified with vertical hydraulic
conductivity less than horizontal
conductivity. We will use MODFLOW to
explore the issue of aquifer stratification
and anisotropy in hydraulic conductivity
using a community well case study.
81
-------
Question 16
Were the assumptions informing the choice and construction of the groundwater flow model
appropriate for the intended use? Please explain.
Reviewer
Name
Reviewer Comment
EPA Response
underlying aquifer. Just NE of well 75m71km there is an old swampy braid
that still contains water and had visible discharge into the Animas River (see
close-up of Google Map image). Also, there is a braid (Coon Creek) about
1200 ft E of well 35m66km.
144) The inclusion of vast areas of rock in the far-field of this model seems
unnecessary and although the model calibration process took care of critical
inputs near the well fields, the largely unknown inputs for the rock and far-
field may draw distracting scrutiny. I did not hear of a sound basis for these
inputs:
horizontal K of the rock = 17100th the horizontal K of alluvium
calling the rock unconfmed with a horizontal base so its saturated
thickness grows from around 30 meters near the alluvial aquifer to
over several hundred meters to the NW (slide 15)
The no-flow boundary around the whole modeled area
The specified head boundaries to the N and W near the model limits
Zero overland flow from the steep rock areas added to the margins of
the alluvium
These poorly-grounded inputs could be avoided by using local scale models
of just the alluvial aquifer, one model for each cluster of wells. The
outermost boundary conditions could be established as specified-head
boundaries where heads of surface waters or wells are known up-gradient in
the alluvium, and by no-flow boundaries along estimated flow lines.
145) Slide 4 shows numerous other wells in the alluvial plain. The domestic
well discharges are not likely to be large enough to have much effect,
especially since most of their discharge returns to the subsurface in leaching
143) We have expanded the modeling
approach (see 134) and created local
models for three of the community wells
to further explore the major areas of
uncertainty, including, 3-dimensional
flow, floodplain heterogeneities and
transient pumping. The regional analytic
element models provide boundary and
initial conditions for the refined local
models. These analyses will be included
in the final report.
144) We have expanded the modeling
approach to include a regional model to
set the boundary conditions for the local
models. The GFLOW model is calibrated
to the observed static water levels at
nearly 300 wells in the nearfield and
baseflow estimated from USGS gages,
giving effective hydraulic conductivity
for the alluvium and surrounding rock,
and an effective regional recharge rate.
The no-flow boundary surrounding the
catchment enforces the water balance. A
local scale MODFLOW model set with
boundary conditions informed by the
regional model is then used to assess
local wells. (Also see response 134.)
82
-------
Question 16
Were the assumptions informing the choice and construction of the groundwater flow model
appropriate for the intended use? Please explain.
Reviewer
Name
Reviewer Comment
EPA Response
fields. Irrigation wells, on the other hand, could have significant discharges,
and some of these are close to municipal wells.
146) Slides 8 and 9 discuss groundwater inflows into the Animas River. The
slides and the notes that were distributed did not explain how this
information was incorporated into the model. Were the modeled groundwater
discharges to the Animas calibrated to values from these analyses? This
should be explained in the text.
147) I'm not sure the analysis of slide 12 is relevant - that applies to cyclic
recharge, and recharge was not included in these models. I have no objection
to omitting recharge, since at the scale of these well-river distances in late
summer, lateral aquifer flows probably far outweigh recharge flows. I also
think that a steady-state analysis is warranted given the limited data, roughly
continuous use of municipal wells, and the close proximity of key wells to
the river.
148) The 2D models presented are capable of only crudely estimating the
fraction of flow that is river water and the timing of river water arrival at the
wells. If there is a great need to quantify simulated concentration vs. time at
a well, a more accurate approach would be to construct a more localized 3D
model that includes all the closer hydrologic boundaries in the alluvial plain.
A 3D flow model with enough vertical discretization to capture the
resistance between the river bed and the well screen could be created with
MODFLOW, AnAqSim, or any number of other 3D groundwater flow
codes. This could be done at one or a few of the wells to quantify the impacts
of the 2D assumption.
145) The steady-state models for the
community wells were based on average
pumping for the period August through
December (non-snowmelt period) and
assumed that the irrigation wells were not
active (non-growing period).
146) We will fully describe the
calibration process and data in the final
report.
147) The updated regional groundwater
(GFLOW) model includes areal recharge
(mountain) as a source of the baseflow in
the Animas River. The local model
(MODFLOW) does not include areal
recharge.
148) We have expanded the modeling
used for this assessment to include
regional flow calibration (GFLOW) and
local groundwater flow at selected wells
(MODFLOW). See (134).
83
-------
Question 17
Were the assumptions informing the capture zone and particle tracking
analysis appropriate for the intended use? How so?
Reviewer Name
Reviewer Comment
EPA Response
Henk Haitjema
149) In general they were, but additional analyses and
discussions are needed as indicated in my answer to questions 15.
Specifically, I have suggested more sensitivity analyses be
conducted, along the lines as to what has been shown on slide 16
and 18. The analysis as presented suggests that the aquifer
responds quickly to transient forcing, allowing for successive
steady state solutions or instantaneous steady state solutions. I
found that analysis to be in error!! In fact, the aquifer probably
responds rather slowly and a steady state model is more suitable
for representing average conditions. In areas (near wells) where
the aquifer may respond neither fast nor slowly, bounding
solutions may be offered assuming both a fast and slow responses
to transient forcing. The issue is discussed in more detail as part
of my answer to question 15.
149) See response (134). The modeling approach
has been expanded to better accommodate aquifer
conditions near rivers, and transient vs steady state
pumping. Both reviewers concluded that there
were considerable uncertainties in applying
WhAEM to model groundwater movement in river
alluvium, especially when trying to detect a short-
term interaction. The expanded modeling approach
performs a floodplain-scale (regional) calibration
using GFLOW and local 3-D modelling at selected
wells using MODFLOW. The modeling will
evaluate the effects of floodplain heterogeneities
and transient forcing. The modeling will be fully
described in the final report.
Charles Fitts
150) As noted under question 16 above, these 2D simulations
probably overestimate the fraction of well discharge coming from
the river, and probably underestimate the travel time a
conservative tracer takes from river to well.
The assumed porosity of 0.20 is probably low for these
sediments. Such values are common in poorly sorted (widely
graded) unconsolidated materials like glacial till, but for alluvial
sands and gravels 0.25 < n < 0.35 is a more common range. The
impact of using a higher porosity in the simulations would be
slower velocities and increased travel times between river and
well.
150) A sensitivity analysis representing a range of
porosities will be included in the final report. We
agree with the suggested values.
84
-------
Question 17
Were the assumptions informing the capture zone and particle tracking
analysis appropriate for the intended use? How so?
Reviewer Name
Reviewer Comment
EPA Response
151) The graphs in slides 25 and 26 would be much more striking
with an arithmetic, not logarithmic Y-axis.
152) The presentation did not show models of the RK 171 or RK
179 wells in NM, so it is not possible to comment on those
analyses and their conclusions. I assume the final report will
provide details on those.
151) The graphs of metals concentrations will be
presented with arithmetic y-axes in the final report.
152) Modeling of a second study area on the lower
Animas River near Aztec, New Mexico was not
completed at the time of the peer review. The
same modeling approach described in (134) will be
applied to this area and will be incorporated into
the final report.
85
-------
Question 18
Did the method for calibration of the local scale groundwater flow model performance to the
observed drawdown reported in the driller's log serve as an effective method? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
Henk Haitjema
153) There was a lack of calibration targets (only one USGS monitoring
well), but that well has been used in this study to arrive at calibrated values
of hydraulic conductivity in the alluvium, see slide 15. In my answer to
question 151 described additional potential calibration targets that should be
explored.
153) The expanded modeling will
include the regional GFLOW model of
the mid Animas River floodplain with
calibration to almost 300 static water
levels as test points (piezometers).
Charles Fitts
154) The key factors in estimating the fraction of well discharge from the
river are aquifer transmissivity (T), ambient hydraulic gradients, well
discharge (Q), and well location relative to the river. The models based input
T and Q values on yield test data and river/well locations on maps, which are
reasonable approaches. The estimation of ambient gradients was probably
the greatest source of uncertainty. Using the USGS well near well near RK
70 was helpful for constraining modeled gradients in that area. However,
there are other data that could help constrain the modeled gradients
elsewhere. The available aerial images and maps should be examined to
determine the location and elevations of other surface waters in the vicinity
of the simulated municipal wells. The modeled background gradients should
be consistent with observed features, and some surface waters may need to
be included as boundary conditions. For example, modeled heads should not
be well above ground surface or above surface water elevations, and surface
water elevations should not be far above simulated heads. This would be
easier to do with smaller, local-scale models.
155) There is uncertainty in many model input parameters (e.g. T
distribution, saturated thickness of the aquifer, well screen elevations,
vertical conductivities), so a more thorough approach would be to run a suite
of simulations (realizations) that investigate the range of possible input
variations. However, if the aim of the modeling is to only determine which
municipal wells have a chance of pumping a significant amount of river
154) We have refined the calibration of
the regional mid Animas River
floodplain model to include the static
water level and baseflow calibration
targets. The regional model provides
boundary conditions to the local models
applied to the selected community
wells.
155) The EPA/ORD team agrees that
the use of the simple models needs
further justification. More complex
local models will be used to address the
uncertainties of modeling groundwater
movement in floodplain alluvium (see
134). The expanded approach will be
included in the final report.
86
-------
Question 18
Did the method for calibration of the local scale groundwater flow model performance to the
observed drawdown reported in the driller's log serve as an effective method? Please explain.
Reviewer Name
Reviewer Comment
EPA Response
water, the WhAEM simulations with just a single realization probably
suffice to draw the broad conclusions shown in the first two "Model Results"
columns in the table on slide 28, except in the case of well 75m-71km,
which was a close call and could use a more sophisticated analysis to draw
conclusions.
87
-------
Part 6: EnviroAtlas Modeling
Question 19
Are the sources of the data included in the maps valid, complete, and adequately documented?
Are there any points of confusion, gaps, or suggestions for improvement?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
156) It seems as though the purpose of the Atlas modeling, and
Atlas itself, were only briefly described in one slide. The
purpose of the modeling and data sources were not entirely
clear. For example, how do the data sources used in Atlas differ
from those in any other publicly available national geospatial
databases, such as USGS databases including their 'National
Map'? Or do they use some of the same data sources? This all
needs more explanation.
156) The EPA/ORD team utilized the
EnviroAtlas tool to access background data as
well as to animate study results, such as the
plume travel and time series of monitoring data
throughout the 600-km reach of river as a
product of analysis. The input data to the
animated map of the movement of the plume
was the WASP modeling results. This
information on data sources will be fully
discussed in the final report.
Charles Fitts
157) I understand that in this section we are to comment on the
animations that were presented. I can't comment on whether
data sources were valid or complete, since only limited
information was included on the animation slides. I think the
slides were reasonably labeled and clear about what they were
presenting. I found both animations to be quite helpful to
visualize the plume's migration and dispersion, and to visualize
the spatial variability of concentration data and the response of
the system to localized precipitation events. I'm sure these
animations would be even more useful to a lay person than a
scientist with more background.
157) The source of the animated data will be
made clearer in the maps and final report. We
appreciate the comments regarding EnviroAtlas
as an animation tool for depicting some of the
analysis results. References to data sources will
be provided in the final report and animated
results will be available to the public on the web.
-------
r
Question 19
Are the sources of the data included in the maps valid, complete, and adequately documented?
Are there any points of confusion, gaps, or suggestions for improvement?
Reviewer Name
Reviewer Comment
EPA Response
William
Stubblefield
158) It is assumed that the sources of the data that are included
in the Atlas modeling maps are the same as those described in
the overview presentation and listed on page 64 of the Empirical
Analyses presentation. No additional explanation for the data
contained in the specific maps was provided therefore it is
difficult to address the validity, completeness, and adequacy of
the documentation.
158) The animated maps show results from the
WASP plume modeling in one instance, and the
spatial and temporal distribution of EPA water
quality samples in other cases. The EPA/ORD
team will ensure that the source of the data
depicted in these products is clearly described in
the final report.
89
-------
Question 20
Do all of the maps and charts communicate the analysis methods and results in such a way as to be readily understood by stakeholders
with interest in the impacts of the Gold King Mine spill (e.g., First Nations; NGO's; news media; and State water, recreation, public
health, and wildlife managers) ? Are there points of confusion, gaps or suggestions for improvement?
Reviewer Name
Reviewer Comment
EPA Response
Brian Caruso
159) No. The maps and charts for the Atlas
modeling need more explanation as to their purpose
and what the results are attempting to show and
what they are actually showing. The usefulness of
these maps is not yet clear.
159). The EPA/ORD team utilized the EnviroAtlas tool to
access ancillary data to help put the area into context and
highlight potential additional sources of metals and potential
exposure. These will be clarified in the final report.
Charles Fitts
160) As noted throughout my comments, there is a
need for editing, clarifying, and polishing. We were
reviewing early draft figures and tables without the
benefit of much additional text. The final report will
have much more text to qualify and explain the
tables and figures, which will be a big help.
160). We are continuing to edit and clarify the background
information slides and the animations. We will explain and
qualify the figures and tables in the final report.
Henk Haitjema
161)1 trust that the ATLAS maps referenced here
are the animations of total metals and arsenic
migration along the Cement Creek, Animas River,
and San Juan River. I felt that these animations were
an effective way of communicating the plume
migration through the system and thus quite
informative for the stakeholders. Other ATLAS
maps that have been used in the various
presentations were also quite helpful in
communicating the spatial and temporal
relationships under discussion. Overall, I find
ATLAS an impressive communication tool.
161) The animations of metals concentrations along the river
systems were provided during the peer review. These products
will be available to the public as part of the report products.
90
-------
Part 7: Bioaccumulation
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
Charles Fitts
162) I am not an expert in this area, but from what I
understood of the discussion, it seems that the
partitioning coefficient approach that was used in this
section may not be appropriate for metals that have
highly regulated concentrations within organisms.
162) BASS does not use chemical partitioning (i.e., BCFs, BAFs
etc.) to simulate the chemical exchanges between a fish and its food,
feces, or gill water. BASS is a differential equation model that
simulates the growth, population dynamics, and bioaccumulation
dynamics of age-structure fish communities. Chemical uptake from
food and gill water and chemical excretion to feces and gill water are
simulated as individual kinetic processes. BASS uses, chemical
partitioning to describe the internal distribution of chemicals
between the water, lipid, and non-lipid organic matter of a fish's
whole body and feces but this is not the level of partitioning
discussed during the review. Also see response to comment (198).
Although aquatic biota certainly regulate their internal metal
concentrations by a variety of physiological processes, these
concentrations are also strongly determined by the diffusion
gradients that determine the excretion and uptake of metals to and
from the organism's ambient water and its food/feces. These later
processes are represented in detail in BASS. BASS, however, also
models the physiological storage and sequestration of metals as
partitioning and complexation reactions between the fish's internal
water and its lipid and nonlipid organic matter. Since the Peer
Review, BASS's algorithms have been updated to accommodate
91
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
saturation kinetics and have been tested using fish data collected in
the Animas River just prior to the GKM release.
The EPA/ORD team will improve presentation of the analytical
system integrated into BASS to clarify how bioaccumulation of
metals is determined.
Kirk
Nordstrom
163) I was considerably puzzled by the presentation on
the BASS code for several reasons. I found the model
that the code is based on to be interesting and
theoretically appropriate, but it seemed to me to require
so much empirical information that it could be decades
before it might be useful. This was confirmed by some
of the discussion because there clearly is some debate
in the scientific literature as to the practical application
of the model. It does seem useful for certain groups of
organic contaminants for which it was originally
designed but not for metals toxicity. More importantly,
I failed to see how this model, with parameters that
would come from longer term experiments than the
lifetime of the GKM plume, had any relevance. The
GKM was a rapid transient event and only similarly
transient experiment data with fish would be similar in
application. Hence, the caged fish study and the lack of
observable fish kills would seem to be the only relevant
information to gage toxicity.
163) We do not agree with the conclusion that BASS requires "so
much empirical information that it could be decades before it might
be useful". Currently, BASS's auxiliary parameterization software
provides users with complete default data sets of bioenergetic and
ecological parameters for 691 species of North American fish. This
software also can configure a default food web for any combination
of these species. The input that must be provided by a user is actually
rather modest. Generally, user supplied data is limited to providing
1) a limited number of chemical properties, 2) time series of
dissolved water concentrations, and 3) physical features of the water
body of interest (e.g., water temperature and depths).
Although BASS outputs can be used to evaluate/gauge potential
residue-based toxicity for fish, BASS was originally developed to
assess dietary exposures to predatory fish, piscivorous wildlife, and
humans for ecological and human health risk assessments; BASS
was not developed to be a toxicity model per se as was the Biotic
Ligand Model (BLM). Unfortunately, this fundamental difference
between the BASS and the BLM was lost during much of the
ensuing discussion during the peer review.
92
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
We agree that observations of toxicity such as the caged fish study
and measurements of body burdens are very important to
understanding the acute toxicity of the GKM plume as it passed
through the system. However, we do not agree that "the lack of
observable fish kills would seem to be the only relevant information
to gage toxicity". As with the other natural processes that were
modeled, the team feels that bioaccumulation modeling adds insights
not possible from fish observations alone. The modeling enables
estimation of potential dietary exposures to wildlife and humans
eating fish from the Animas River following the GKM release and
evaluates the assumption that the plume passed too quickly for fish
to have taken up metals. To this end, we used BASS to the
bioaccumulation of As, Cd, Cu, Pb, and Zn before, during, and after
the GKM plumes at Silverton and Durango based on BASS's tested
and peer reviewed status with respect to methylmercury, which also
binds to fish proteins containing sulfhydryl groups.
93
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
William 164) As stated on the EPA web site, the BASS model
Stubblefield was developed to "predict the population and
bioaccumulation dynamics of age-structured fish
assemblages exposed to hydrophobic organic
pollutants and class B and borderline metals that
complex with sulfhydryl groups (e.g., cadmium,
copper, lead, mercury, nickel, silver, and zinc)." In the
scope of the GKM effort, the BASS model was used
to: 1) predict tissue metals concentrations in trout
resulting from estimated dissolved metal
concentrations in the Animas and San Juan Rivers and
2) using the estimated tissue concentrations predicted
from BASS, make an assessment of "short-term
impacts" to trout populations by comparing these
values to residue-based tissue concentrations reported
in the review by Jarvinen and Ankley (1999).
However, this approach seems to be somewhat lacking
for a number of reasons.
165) First, the BASS model was developed to predict
tissue concentrations using a BCF/BAF approach. This
assumes that the concentration of a chemical in the
tissues of exposed organisms is a function of
waterborne (or waterborne and food) concentrations
and that the uptake of the material into tissues is a
function of exposure concentration. Steady state
concentrations are reached when the rate of uptake is
164) The EPA/ORD team agrees with this description of goals and
objectives of the bioaccumulation modeling are what was intended.
165) There is a significant difference between a BCF/BAF approach
and the bioaccumulation modeling used in this study. The
BCF/BAF approach to predicting chemical concentrations in
aquatic biota assumes that the ratio of an organism's whole body
chemical concentration [C0(0] to the freely dissolved concentration
of that chemical [Cw(0] at any time t is a constant (i.e., BCF if the
organism accumulates the chemical only from water, or BAF if it
accumulates the chemical from both water and food).
94
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
equivalent to the rate of elimination. Most, if not all, of
the materials that the model has been used with in the
past are neutral, lipophilic organic compounds (e.g.,
DDT, PCB) or organometallic compounds (e.g.,
methyl-mercury) that follow these kinetics (i.e.,
concentration dependent kinetics). Metals, on the other
hand, do not follow this model. Many metals are
"essential" for life processes (e.g., Co, Cu, Zn) and
their concentrations in the body are homeostatically
controlled to maintain "constant" concentrations
necessary for life processes. BCF values are calculated
as the quotient of the internal tissue concentrations
divided by the exposure concentration. Thus, with a
metal when external concentrations are low, the body
actively "concentrates" metals to maintain necessary
internal concentrations resulting in extremely high
BCFs; in situations when metals exposure
concentrations are elevated but tissue concentrations
are maintained at homeostatic levels, BCFs are low. A
recent review by DeForest et al (2007) states that
"Results indicate that field BAFs, like laboratory
BCFs, tend to be significantly (p<0.05) inversely
related to exposure concentration" and "Data
presented indicate that for metals and metalloids,
unlike organic substances, no one BAF or TTF can be
used to express bioaccumulation and/or trophic
transfer without consideration of the exposure
Mathematically, this relationship is described by the following
simple differential equation
dt
dt
dt
This is the simplest model used to describe the bioconcentration or
bioaccumulation of chemicals in aquatic organisms and is most
commonly used only for screening level risk assessments for which
time dynamics can be safely ignored. BASS, however, does not use
this approach to simulate chemical concentrations in fish. Rather,
BASS simulates these concentrations by solving the following
system of equations
B
wf
dBf
~dT
dWf
= F-E-R-EX-SDA
where
Cf is the fish's chemical concentration;
Bf is the fish's chemical body burden;
Jg is the net chemical exchange across the fish's gills to and
from water;
95
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
concentration." McGeer et al (2003) conclude from
their review on the topic of BCFs that:
"The accumulation of Zn, Cd, Cu, Pb, Ni, and
Ag in aquatic biota were, in general,
remarkably consistent, particularly for Zn,
where total body/tissue concentration varied
little over a wide range of exposure
concentrations, exposure conditions, and
species. However, mean BCF values for the six
metals were characterized by high variability,
and there was an inverse relationship between
BCF and exposure concentration. Therefore,
using the weight of evidence available, it is
virtually impossible to derive a meaningful
BCF value that one could say is representative
of the BCF for each of the metals. Even when
BCFs are limited to the exposure range where
chronic toxicity might be expected (based on
water-quality guidelines), it is not possible to
derive a precise and accurate BCF value
To correctly assess potential hazards, it would
be necessary to distinguish between essential
nutritional accumulations, that which is
sequestering and stored, and accumulation that
causes adverse effects. Because BCFs are
Ji is the net exchange across the fish's intestine to and from
food/feces;
Jm is the chemical's rate of biotransformation;
Wf is the fish's body weight;
F, E, R, EX, and SDA are fish's rates of feeding, egestion,
respiration, excretion, and specific dynamic action,
respectively.
The reviewer noted that several recent papers demonstrate that
BCFs and BAFs are generally inversely related to exposure
concentrations, and questioned BASS's ability to reproduce such
relationships. A detailed analysis of the time varying BAFs [i.e.,
BAF(0=C/(0/CW(0 where Cf(t) and Cw(f) denote the fish's whole
body concentration and dissolved water concentration at time t,
respectively] predicted by BASS for the Silverton and Durango
plumes demonstrate that BASS indeed predicts an inverse
relationship between BAF(r) and Cw(f). (See the inserted figure
below.)
For example, the slope of the simple linear regression log
BAF(0=a+b*log Cw (t) for the Silverton plume was -1.01 [P(b=0)
<10"6; n=1009]. Similarly, the slope of the same regression model
for the Durango plume was -1.08 [P(b=0) <10'6; n=2663]. Similar
results were obtained for Cd, Cu, and Pb at both Silverton and
Durango. For arsenic, however, only the Silverton demonstrated a
strong inverse relationship between BAF(^) and Cw (t) since there
was no arsenic detected as the plume passed Durango.
96
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
based on the whole-body concentration, the
BCF model does not distinguish between these
different forms of bioaccumulation and
therefore it would seem unlikely that the
criterion would be correlated to adverse effects
such as chronic toxicity."
Silverton and Durango BAF Analysis
10000
1000
u
c
M
100
i
0.01 0.1 1
Zinc dissolved water cone (mg/L)
10
The EPA/ORD team agrees with the concerns about using a BCF or
BAF approach to assess metal bioaccumulation. The problems
associated with attempting to use a single BCF of BAF without
reference to the fish's aqueous exposure conditions are real and not
restricted to metals; the same problems can be manifested by
hydrophobic organic chemicals, especially those having "high" log
Kow. To overcome this problem, we used BASS to simulate the
bioaccumulation dynamics of metals in Animas River fish before,
during, and after the GKM release as a part of the exposure
analysis. The approach will be fully documented in the final report
and will be subject to another peer review at that time.
97
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
166) The second major concern with the approach
employed stems from the estimation of adverse effects
based on tissue concentrations (i.e., based on
comparison of tissue residue levels to those reported
by Jarvinen and Ankley (1999)). This approached is
built upon the Critical Body Residue (CBR) concept.
CBR is the concentration of chemical bioaccumulation
in an aquatic organism that corresponds to a defined
measure of toxicity (e.g., mortality, reproductive
impairment). Rainbow, in his 2002 article, "Trace
Metal Concentrations in Aquatic Invertebrates: Why
and So What?" concludes: "Toxicity is related to a
threshold concentration of metabolically available
metal and not to total accumulated metal
concentration." Finally, Adams et al. 2010, concluded
that "Available information suggests that it is not
possible to develop universally applicable whole-body
CBRs for metals (except for Se, methylmercury or
other organo metals[sic]). Aquatic organisms
differentially handle accumulated metals with respect
to storage, detoxification, and excretion. As a result,
measuring total metals in an organism provides limited
information on the metal concentration associated with
the biologically active pool. However, the benefits of
monitoring for contaminant trend and exposure
assessment are acknowledged."
166) The residue-based toxicity approach is based on the simple
assumption that chemical concentrations which are internal to an
organism (e.g., critical body residues, CBRs) are inherently more
accurate for assessing onset of toxic responses in organisms than
are external environmental concentrations or benchmarks (e.g.,
LCso, etc.). Although we firmly believe in the residue-based toxicity
approach, we do not assert that there is one CBR or CBR-like
threshold for each metal and fish species. Rather, depending on the
endpoint of concern, there is a range of CBRs which are useful in
making objective and well defined ecological risk assessments.
These CBRs, however, will vary not only by the species of concern
but also on that species life history (i.e., age, body weight, trophic
position etc.) and exposure history. We acknowledge that it is
important to discuss and quantify the uncertainties/variabilities of
using this approach and we will do so in the final report.
For this study, we simply assumed that if the range of BASS-
predicted whole body metal concentrations of Animas River fish
during and immediately after the GKM release plumes significantly
overlapped with the range of CBRs reported by Jarvinen and
Ankley (1999), then further analysis and study could be warranted.
The BLM is a residue-based toxicity model that uses metal
concentrations in the fish gill to assess acute metal toxicity. During
the Peer Review, the reviewer strongly argued for its use in
assessing the acute impacts of the GKM release. Although the
EPA/ORD team is certainly open to investigating the utility of the
98
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
167) Finally, based on the presentations made it seems
clear that the waterborne exposures following the Gold
King Mine incident were reasonably short-term in
nature (hours not days) and were characterized by an
initial spike in concentration that dissipated rapidly
returning to pre-spill conditions. Kinetics of such an
exposure would suggest that steady-state whole body
tissue concentrations would not have been achieved
given the duration of the exposure and its variable
nature. Initial impacts to organisms would likely have
been acute in nature due to the initial pulse exposure.
Whole body tissue concentrations would not reflect
possible effects to organisms.
168) If the analysis objective of the bioaccumulation
and residue-based effects evaluation was to "assess the
expected implications of the Gold King Mine release
on Animus River biota" it would seem that a more
traditional and straightforward approach to evaluating
the potential impacts could be achieved by comparing
estimated exposure concentrations for individual
BLM, we also remain committed to the whole body CBR approach
since whole body and tissue concentrations of metals and organic
chemicals in fish should be significantly correlated with one
another. This assertion is certainly true for whole body and fillet
concentrations (e.g., Bevelhimer, M. S., et al. 1997. Estimation of
Whole-Fish Contaminant Concentrations from Fish Fillet Data. Oak
Ridge, TN, Oak Ridge National Laboratory), whole body and liver
concentrations (e.g., Goldstein and DeWeese 1999, JAWRA
35:1133-1140), and fillet and blood concentrations (e.g., Schmitt et
al. 2009, Arch. Environ. Contam. Toxicol. 56:509-524).
167) The goal of the modeling was to investigate whether the
assumption expressed by the reviewer would be supported by a
bioaccumulation kinetics analysis of exposure to the plume
concentrations. We did not start with this assumption but set out to
investigate it. The very high concentrations that characterized the
plume lasted for hours but the full plume duration was more like 48
hours at the downstream locations. We will assess exposure to
acute levels of metals defined by water quality criteria. However,
the EPA/ORD team feels that the analysis adds additional insights
gained from comparing metals and can be also be used to discuss
potential human dietary exposure. We note that the bioaccumulation
modeling suggested that metals would have been taken up by trout
during the relatively short exposure time and also expelled quickly.
The modeling also suggested that critical levels were not achieved
and mortality would not be expected given this rapid uptake and
depuration, consistent with the reviewer's general interpretations.
99
-------
Question 21
Given the limitations of the BASS model, how appropriate is the simulation of
bioaccumulation of As, Cd, Cu, Pb, andZn in theAnimas River trout fishery?
What are the strengths and weaknesses of using this approach?
Reviewer
Name
Reviewer Comment
EPA Response
metals to appropriate state standards or US EPA
Ambient Water Quality Criteria.
References:
Adams WJ, Blust R, Borgmann U, Brix KV, DeForest DK, Green
AS, Meyer JS, McGeer JC, Paquin P, Rainbow PS, Wood CM.
2010. Utility of Tissue Residues for Predicting Effects of Metals
on Aquatic Organisms. Integrated Environmental Assessment and
Management. 7:75-98.
DeForest DK, Brix KV, Adams WJ. 2007. Assessing metal
bioaccumulation in aquatic environments: The inverse
relationship between bioaccumulation factors, trophic transfer
factors and exposure concentration. Aquatic Toxicology 84:236-
246.
Jarvinen AW, Ankley GT. 1999. Linkage of effects to tissue
residues: Development of a comprehensive database for aquatic
organisms exposed to inorganic and organic chemicals.: Society
of Environmental Toxicology and Chemistry Pensacola, FL
McGeer JC, Brix KV, Skeaff JM, Deforest DK, Brigham SI,
Adams WJ, Green A. 2003. Inverse Relationship Between
Bioconcentration Factor and Exposure Concentration for Metals:
Implications for Hazard Assessment of Metals in the Aquatic
Environment. Environmental Toxicology and Chemistry,
22:1017-1037
Rainbow PS. 2002. Trace metal concentrations in aquatic
invertebrates: why and so what? Environmental Pollution
120:497-507
Copper concentration in 0.6-0.8 kg Brown Trout
(0.07
-------
References
Versar, 2016. Peer review of EPA's Gold King Mine analysis of fate and transport in the
Animas and San Juan Rivers. Versar. Inc. 6850 Versar Center, Springfield, VA 22151.
-------
&EPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
Recycled/Recyclable Printed on paper that contains a minimum of
50% postconsumer fiber content processed chlorine free
------- |