United States
  Environmental Protection
                                            October 2014
Responses to External Peer-Review Comments
   Risk Assessment of Spent Foundry Sands
           In Soil-Related Applications
           U.S. EPA Office of Resource Conservation and Recovery
                Economics and Risk Assessment Staff

         U.S. Department of Agriculture-Agricultural Research Service

                    The Ohio State University
T • H  • E

[This page intentionally left blank.]

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications

                                Table of Contents

1)  Introduction	1
2)  Summary by Reviewers	2
3)  Characterization of Foundry Sands	4
    a) General Sand Properties	4
    b) Beneficial Use Characterization	5
    c) Management of Foundry Sands	5
    d) Representative Sampling	6
    e) Constituents Sampled	7
    f) Detection Limits	8
    g) Documentation	8
    h) Additional Sources of Information	9
4)  Problem Formulation	10
    a) Organizational Issues	10
    b) Selection of Constituents to Model	12
    c) Highly Exposed/Sensitive Subpopulations	13
    d) Storage Pile Conceptual Model	15
    e) Roadway Construction	17
    f) Dermal Exposure	18
    g) Manufactured Soil Conceptual Model	19
5)  Screening and Modeling	22
    a) Positive Comments	23
    b) Soil-Blending Site Distances	24
    c) Groundwater Model	25
    d) Consumption Model Protectiveness	29
    e) Constituents Contributed by Soil	30
    f) Unitized Exposure Estimates	30
    g) Transparency Issues	31
6)  Risk Characterization and Uncertainties	33
    a) Findings/Conclusions	33
    b) Soil Properties, Background, Phytotoxicity, and Soil Biota	35
    c) Sensitivity Analysis	36
    d) Variability versus Uncertainty	36
    e) Inconsistencies with the Exposure Factors Handbook	39
    f) Child-Specific Exposure Factors Handbook Not Used	41
    g) Data Collection Uncertainty	41
    h) Toxicity Value Uncertainty	43
    i) Consumption Rate Uncertainty	43
    j) Cumulative Risk	47
    k) Clarity	51

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

7)  General/Other	52
    a) Non-Technical Abstract and Public Label	52
    b) Technical Inaccuracies/Editorial Comments	53
    c) Application to States	54
    d) Risk Assessment versus Risk Management	55
8)  References	56
Appendix A) IWEM modeling review, alternate model search, and recommendation for
evaluating the SFS home garden scenario groundwater pathway	1
    Appendix A: References	16

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
1) Introduction

In 2002, the U.S. Department of Agriculture's Agricultural Research Service (USDA-ARS)
implemented the Foundry Sand Initiative to evaluate the reuse of spent foundry sands (SFS) in
horticultural and agricultural applications. As part of this effort, the U.S. Environmental
Protection Agency (EPA) worked in collaboration with USDA-ARS and Ohio State University
(hereafter collectively referred to as "the Authors") to investigate the potential risks associated
with such activities, and produced, through contract with RTI International (RTI), a draft report
entitled "Risk Evaluation of Spent Foundry Sands in Soil-Related Applications" (Risk

Subsequently, the Authors retained Industrial Economics, Incorporated (ffic) to conduct an
independent peer review of the Risk Assessment.

The review panel was charged with providing comments on the following:
    1.  Please comment on the transparency of the risk assessment.
    2.  Please discuss the adequacy of the risk assessment execution.
    3.  Please comment on whether the selection of U.S. foundries was representative of the
       industry and if the characterization of these foundry sands was adequate.
    4.  Please comment on the methodology used for choosing constituents to evaluate.
    5.  Please comment on the conceptual models, particularly the plausibility of the sources,
       pathways, and receptors included.
    6.  Please discuss the appropriateness of the Manufactured Soil conceptual model, as
       protective of the other conceptual models.
    7.  Please discuss whether the screening steps reported in Chapter 4 were appropriately
       conservative in their application to support the conclusions.
    8.  Please comment on the appropriateness of the various probabilistic modeling steps
       employed to develop national-scale screening values.
    9.  Within the context of a screening risk assessment, please comment on the level of
       conservatism inherent in the Home Gardener scenario, with special attention to the
       assumption of independence of the ingestion pathways. Please also comment on the
       rationale for modeling the 50%tile and 90%tile general population consumption rates,
       each with a 50% homegrown fraction.
    10. Please comment on how soil background, phytotoxicity, and impacts on soil biota were
       considered in the assessment.
    11. Please comment on the clarity of the Risk Characterization section, with special attention
       to the discussion of uncertainties.
    12. Please comment on whether the assessment supports the report's conclusions.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
When identifying and selecting experts for the review panel, lEc made an effort to include
individuals with expertise in one or more of the areas outlined in Table 1.

                 Table 1: Areas of expertise sought in potential peer reviewers
Area of Expertise
Human Health Risk Assessment
Spoil/Plant Science
Groundwater Hydrology
Expertise in the methods and approaches to conducting human
health exposure and risk assessments, including experience
creating or reviewing exposure and risk assessment documents
and familiarity with multimedia risk assessment.
Expertise in the field of soil science, including metals transport
in soils and metals uptake in plants.
Expertise in the methods and approaches used for modeling the
fate and transport of contaminants in groundwater, as well as the
effects of soil properties on groundwater movement
The final panel of expert reviewers included (with area of expertise in parentheses):

   •   Dr. Ken Barbarick, Colorado State University (Soil Science)
   •   Dr. Mary Fox, Johns Hopkins University (Human Health Risk Assessment)
   •   Dr. Charles Harvey, Massachusetts Institute of Technology (Groundwater Hydrology)
   •   Dr. Donna Vorhees, The Science Collaborative (Human Health Risk Assessment)

This comment review and response report focuses on the technical themes presented by the peer
reviewers of the SFS risk assessment. Section 2 provides each reviewer's summary, while
Sections 3 through 13 address specific reviewer comments. For each comment category, the
comments are provided by reviewer, followed by the response.

2) Summary  by Reviewers

   Dr. Ken Barbarick

       / think the report did do a comprehensive risk assessment of the use of spent (recycled)
       foundry sands. I support their conclusions that their "Home Garden" scenario is
       protective of human health. I recommend that they include leaching of constituents for
       the storage pile as a part of the modeling process and that they pursue microbial-
       community studies to better characterize the impact on soil biota. I do not believe that
       this report has answered all necessary questions (i.e., the impact on specific soil biota).
       Several more studies would be needed to also quell the concerns expressed by the
       Michigan Department of Environmental Quality. I would characterize the report as an
       excellent start and foundation, but it is not a complete vetting of the potential impacts.

   Dr. Mary Fox

       Overall approach of screening steps leading to a more refined analysis of selected
       constituents is sound. Parts of the report are poorly organized and lack clarity,
       particularly sections of Chapters 3 and 6 and the rationale and approach to the
       probabilistic modeling. There are problems with implementation of the probabilistic

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       modeling that compromise the conservatism of the home-produce ingestion pathway
       and, ultimately, the risk assessment findings and conclusions.

       *   Data inputs for ingestion scenarios (particularly home gardener) must be double
          checked for accuracy and revised to reflect source data in some cases.
       *   Probabilistic analysis of soil and produce ingestion scenarios must be revised and
          repeated before concluding that use of manufactured soils will be protective of
          human and ecological receptors.

   Dr. Charles Harvey

       This risk assessment synthesizes a remarkably wide range of data and models of
       environmental processes. The breadth of the study is impressive, and the assessment
       makes ingenious use of a variety of existing models and data sets. As an academic
       researcher, it is easy to suggest that some data sets are insufficient to fully characterize
       the range of conditions across the United States and that we do not yet understand
       some physical and biogeochemical processes well enough to construct accurate
       models. However,  many of these complaints would be counterproductive - the point of
       this assessment must be to construct the best estimates of risks given available data
       and existing models. Therefore,  I will focus this review primarily on basic conceptual
       issues and on aspects of the evaluation that can be improved with available methods
       and data. I have chosen to first construct a list of broad comments and to note how
       these comments relate to the charge questions. I then provide a few specific questions,
       and finally to come back to the charge questions  with specific responses.

   Dr. Donna Vorhees

       This report benefits from recent research targeting spent foundry sands (SFS)
       characterization that supports the evaluation of exposure.  SFS appears to contain
       chemical concentrations that are similar to what is found in undisturbed soils under
       natural conditions. In addition, the Authors explore whether other factors besides
       concentration might result in more exposure to chemicals in SFS relative to natural soil
       and conclude that there is limited potential for such increased exposure.  Therefore, it is
       understandable to propose use of SFS for manufactured soil and other beneficial uses.
       The Authors present a labor-intensive, national-scale risk assessment to determine risk
       associated with likely uses of SFS. Like all risk assessments, this one is inherently
       uncertain.  The Authors understand this reality and carefully explain many sources of
       uncertainty both qualitatively and quantitatively in the form of a probabilistic risk
       assessment for exposure pathways believed to be associated with the highest levels of
       exposure. But as has been noted by EPA Region 9 in its comments, this risk
       assessment might serve as a model for similar assessments of materials that might pose
       greater risk than SFS to ecological and human health. Therefore, the assessment should
       be viewed in this light and held to a high standard with respect to methodology and
       documentation. In general, my comments focus on opportunities to improve what is
       generally a sound and useful risk characterization of the beneficial use of SFS.

Major issues identified by the peer reviewers requiring attention are the following:

       1.  The risk assessment should incorporate exposure information from EPA's Final
          Child-Specific Exposure Factors Handbook (September 2008). This document
          provides more recent reviews of exposure data and recommendations for point

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
          estimate and distributions of some risk model inputs than those provided in EPA's
          1997 Exposure Factors Handbook.
       2.  The probabilistic risk assessment (PRA) complies with much of what EPA
          recommends in its PRA guidance (U.S. EPA, 2001), but some additional
          documentation and potentially additional analysis is warranted. I refer to EPA
          recommendations in my responses to charge questions.
       3.  The risk assessment should provide a brief, succinct explanation of why, despite
          multiple screening steps, cumulative risks associated with SFS use are below pre-
          established levels of concern.

3) Characterization of Foundry Sands

The peer-review comments relating to the initial characterization of foundry sands are  grouped
into eight subcategories:
   a)  General Sand Properties;
   b)  Beneficial Use Characterization;
   c)  Management of Foundry Sands;
   d)  Representative Sampling;
   e)  Constituents Sampled;
   f)  Detection Limits;
   g)  Documentation; and
   h)  Additional  Sources of Information.

   a)  General Sand Properties

   Dr. Charles Harvey
       Page 1-1, paragraph 2
       Why do heat and abrasion render sands unsuitable? To develop a conception of SFS, it
       would be useful to better understand how it has been altered in the foundry from natural
       sands so that it is no longer useful.

Aside from an initial cleaning process, the foundries themselves do nothing to the native sands
other than add a variety of materials (e.g., clay, seacoal, binders) to make them suitable for
casting. During the casting process, however, the sands are exposed to high temperatures, which
cause the grains to fracture.  In addition, during mold formation and sand reclamation, the sand
grains become abraded as they rub against each other. The fracturing and abrasion ultimately
change the grain shape, which makes the sands undesirable for continued casting. A change in
the grain shape will prevent the gases to pass through the mold and cause it to crack.

The third sentence of the second paragraph on page  1-1 now reads as follows:
       "However, mechanical abrasion during the mold-making process and sand
       reclamation, and exposure to high casting temperatures causes the sand grains to
       eventually fracture. The fracturing changes the shape of the sand grains, rendering

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       them unsuitable for continued use in the foundry. The resulting residuals are
       generally managed as a waste or beneficially used."

   b) Beneficial Use Characterization

   Dr. Mary Fox

       Section 5.3, page 5-5, Probabilistic Modeling of Soil/Produce Ingestion Pathway
       Need more explanation for risk assessment approach to soil/produce ingestion pathway.
       Why not directly estimate risks using the available SFS data? Is it possible to make
       some experimental manufactured soils to develop data for this part of the risk

   Dr. Donna Vorhees

       The Authors should describe past and ongoing use of SFS in more detail. They refer to
       these uses: "Approximately 25% of the 10 million tons of SFS produced annually are
       beneficially used outside of the foundry, but only 3.9% of SFS is used in soil-related
       applications (AFS Survey, 2008)..." The Authors explain in Section 2.4 that they
       conducted a peer-reviewed literature search regarding metals and organics in SFS.  In
       Section 2.5, they refer to a literature search for field studies of SFS leaching.  But they
       found no field studies related to past or ongoing uses of SFS in amended soil. Beyond
       the scientific literature,  have there been reports by other credible sources of any
       problems that have arisen from past or ongoing use of SFS in amended soils?


The Authors found no reports from credible sources of problems arising from past or  ongoing
use of SFS. The report language has been modified to relate this more explicitly. Also, although
geotechnical applications utilize the most foundry sand in the United States, at the time of this
evaluation only one company was identified that used SFS in blending operations. When the
Authors contacted this company (in Ohio), the company had not reported any problems to us
(e.g., plant growth issues). Essentially, the spent sand is used as a replacement for sands that they
would normally purchase for soil blending. Since the constituent concentrations in spent sands
are similar to those of native sands, one would not expect any problems.

The Authors have revised the discussion on the probabilistic modeling of the soil/produce
ingestion pathway to better communicate the risk modeling approach. The intent of this
evaluation was to develop risk-based screening criteria for soil-related beneficial uses of SFS,
rather than to perform a classic risk assessment. Creating/experimenting with manufactured soil
was beyond the scope of this evaluation.

   c) Management of Foundry Sands

   Dr. Charles Harvey

       Page ES-2, paragraph 6
       Are the heavily contaminated sands used for brass or olivine sands ever mixed with
       other sands? In other words, is the distinction between the contaminated SFS left out of

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       this assessment and the safer SFS retained for the assessment always clear? Would
       foundries ever shift from one kind of sand to another and in the process mix the sands?

       Page 2-4, paragraph 3

       Core butts were removed by sieving. Will these butts be removed before the use of SFS,
       and if not, could they be a source of contamination in SFS neglected by this


While most foundries only pour one type of metal, some foundries may pour both ferrous and
non-ferrous metals on separate lines. In the latter case, it is possible for a foundry to mix these
waste streams. The report carefully stipulates which waste streams do and do not qualify as SFS
(e.g., spent molding and core sands qualify, whereas broken or unused cores do not), and which
foundry and sand types were being evaluated. Although the Authors did sample from non-leaded
brass foundries and those that use olivine sand, no mixed waste streams were sampled.

During the casting process, a portion of the core will generally break down to individual grains.
This is because the binder thermally decomposes when it comes in contact with the molten
metal. But in most cases, the cores do not break down completely and residual  pieces remain. In
any case, the resulting spent sands contain sand from both molds and cores, but a much smaller
portion from cores. It is only those core pieces that do not break down that are  removed prior to
soil blending. Therefore, the spent sands analyzed for this risk assessment do consider
contaminants from the cores because individual grains from the broken cores do mix with the
molding sands.

    d)  Representative Sampling

    Dr. Ken Barbarick

       The selection of the foundries and the characterizations of the foundry sands are
       adequate.  The study used a good distribution of geographical and process-types.

       The researchers appropriately eliminated olivine sands for testing since they most likely
       would not be used in a soil mix that grows vegetables or fruits.

    Dr. Mary Fox

       Information provided in the risk assessment is not adequate to evaluate whether
       selection of SFS for analysis was representative of the industry. What is the size of the
       industry? The 43 samples available represent what percent of the industry? Also, the
       sands analyzed were from foundries in the east, south, and mid-west. No samples were
       taken from western states. If geographic representativeness is not relevant to developing
       a national assessment, a justification should be provided.

    Dr. Charles Harvey

       / have no experience with variability among foundries. However, 43 samples appears to
       be a small sample size. More samples would benefit the assessment given the
       differences among natural sands, the metals cast, and the different binders.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   Dr. Donna Vorhees

       The selection of U. S. foundries appears to be representative of the industry. The
       characterization of foundry sands appears to have included chemicals that might
       reasonably occur in SFS and have the potential for causing adverse health outcomes.
       However, additional documentation is needed to ensure the representativeness of the
       foundry sampling effort, which is the basis of all risk analyses.

       The USD A study appears to have collected a reasonably representative sample of SFS
       material over multiple years. However, not until Section 6.8.2 (point 1) do the Authors
       clearly describe how the initial set of foundries was selected for sampling. This
       description aside, industry representatives, rather than USDA scientists, collected
       samples in years 2 and 3, and only a subset of facilities with data for the first year
       provided data in subsequent years. Might there have been any selection bias such that
       facilities with higher chemical concentrations in SFS elected not to report these results or
       did not provide samples at all? Also, is there any reason to think that foundry operations
       might be modified in the future in a way that influences SFS properties? I suspect that
       the Authors considered these questions,  but it would be helpful to document this
       information to provide assurance that this assessment provides an upper bound on
       potential risks under current and future conditions.


The majority of the foundries in the United States are located east of the Mississippi River and in
the Midwest. Therefore, the focus was on foundries from this portion of the country. In addition,
every effort was made when selecting the foundries to cover the widest range of sands that were
amenable for beneficial use in  soil-related applications. Both large and small foundries were
targeted, along with those using a variety of molding and core sand operations.  Green sands were
targeted the most since they represent 80% of the metalcasting byproduct volume in the United
States and are by far the most logical choice for use in manufactured soils. In this assessment,
83% of the foundries targeted for study used green sands, while the remaining used chemically
bonded molding sands. These numbers are representative of industry averages.

   e) Constituents Sampled

   Dr. Mary Fox

       / believe the SFS samples used were adequately characterized. It was helpful to have
       the SPLP leachate data to supplement the TCLP.

   Dr. Charles Harvey

       The list of metals contains the most likely contaminants except for the neglect of mercury
       and selenium in the screening model. I do not have the background to comment on
       potential organic contaminants. However, it is clearly very important to carefully consider
       all possible organic contaminants, and I hope that one of the other reviewers can bring
       this expertise to the review.


Mercury was not included in the data set because the Authors did not have the analytical
capability at the time, but it should not preclude its inclusion in this assessment. As discussed in


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
report section 4.2, a study by Fahnline and Regan (1995) found that mercury concentrations in
TCLP leachates from 52 spent sands were very low at < 0.10 mg L"1.

No leaching data were available for selenium. However, the Authors wish to point out that the
maximum porewater concentration for pure SFS (0.039 mg L"1, Table B-26) is below both the
MCL and tapwater screening level for selenium (0.05 mg L"1 and 0.078 mg L"1, respectively). .
With respect to the organic contaminants, the Authors considered those organics that would
likely be generated as a result of the high-temperature casting process (i.e., EPA priority
phenolics, PAHs, and dioxins).

   f)  Detection Limits

   Dr. Charles Harvey
       Page 2-5, paragraph 3
       "The method detection limit for this data set was calculated by multiplying the standard
       deviation of the baseline noise by the t-value at the 99% confidence interval." This
       statement needs explanation and description of the implications.

This particular statement was misplaced and has been moved in the report to Section 2.4.2, PAHs
and Phenolics. To clarify, the MDL was calculated by multiplying the standard deviation of six
replicate standards by the Student's lvalue at the 99% confidence interval. Calculating the MDL
at the 99% confidence interval allows for  the probability that 1% of the samples analyzed, which
have a true concentration at the MDL, will be false positives. Additionally, reporting data down
to the MDL does nothing to control the possibility for false negatives. The Authors have revised
the text in the report as follows, to improve clarity:

       "The method detection limit (MDL) for this data set was calculated by
       multiplying the standard deviation of replicate standards (n = 6) by the  Student's
       lvalue at the 99% confidence interval. Calculating the MDL at the 99%
       confidence interval allows for the  probability that 1% of the samples analyzed,
       which have a true concentration at the MDL, will be false positives."

   g)  Documentation

   Dr. Donna Vorhees
       Ideally, the final version of this risk assessment will include either (1) the Dayton et al.
       paper with SFS data, which is referred to as "under review," or (2) reference to the
       version of this paper that is accepted for publication in a peer-reviewed journal.  How well
       the data represent SFS is a separate question that is addressed in response to charge
       question #3.

       Sections 1.1 and 1.2 refer to a multiyear research project conducted to characterize
       inorganic and organic constituents in SFS and to assess the potential mobility and
       uptake of these constituents by environmental receptors. The Authors  say that the
       results of this research are in the public domain, but do not list any citations. Including

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

       them all in one place early in the document would briefly convey the scope of the
       multiyear study.

The Authors have added a complete list of relevant citations to Section 1.1., those being Dungan
2006; Dungan and Dees, 2006, 2007, and 2009; Dungan and Reeves, 2005 and 2007; Dungan et
al., 2006 and 2009; and  Dayton et al., 2010.

   h) Additional Sources of Information

   Dr. Donna Vorhees
       In the first paragraph of Section 2.4, the Authors explain that "the foundry industry
       routinely analyzes their sands for metals and/or organics," but these data were not
       considered in the assessment because of "inconsistencies between foundries in the
       sampling and testing protocols." They also refer to a database compiled by Dr. Tikalsky
       at The Pennsylvania State University, but did not use these data either because "method
       detection limits varied for similar constituents, and as a result, comparisons could not
       easily be made between the data." None of the reasons for excluding these data
       suggests data quality problems, just inconsistencies in how data were collected.
       Consequently, it is not apparent why the data were completely dismissed. For example,
       despite differences in sampling and analytical methods, do the data suggest that much
       higher or lower concentrations of any chemicals were found in the other data sets? Might
       there be additional COCs? Or are the data sets generally consistent with the USDA
       data, after taking into account the inconsistencies?

The first statement about foundries testing their sands has been removed from the document, as  it
is somewhat irrelevant. While many foundries do analyze their sands for waste disposal and
beneficial use requirements, the Authors did not have access to these data for the purposes of this
risk assessment.

With respect to Dr. Tikalsky's database, it should be clarified that it examined 338 foundry sand
byproducts, not just spent molding sands. The database did not analyze any additional
constituents of concern that were not  addressed in the assessment of the  43 spent sands. In fact,
the Tikalsky database included very little overall information about the constituent
concentrations, and this information was not consistent between the sands. While many of the
sands in the Tikalsky database were not suitable for  beneficial use in soil-related applications
(e.g., chemically bonded sands, shot blast fines), the concentration of metals (total and teachable)
and organics in iron, steel, and aluminum moldings  sands were comparable with the numbers
the Authors obtained when analyzing the 43 spent sands for this assessment. This section of the
report has been rewritten as follows to note the similarities between the constituent
concentrations in the Tikalsky database and the 43 sands used for the assessment:
       "A database was created by The Pennsylvania State University (Penn State),
       where industry data on different foundry waste materials were compiled (Tikalsky
       et al., 2004). This database contains interesting information on total and teachable
       concentrations of various constituents in foundry byproducts, many of which were
       not suitable for beneficial use in soil-related  applications. While  the Penn State

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       database was not used in this risk evaluation as a result of inconsistent analytical
       data among the foundry byproducts, a preliminary comparison of the database
       with the USDA data set revealed that metal (total and teachable) and organic
       concentrations in molding sands were highly similar."

4) Problem Formulation

The peer-review comments relating to up-front problem formulation are grouped into seven

   a)  Organizational Issues;
   b)  Selection of Constituents to Model;
   c)  Highly Exposed/Sensitive Subpopulations;
   d)  Storage Pile Conceptual Model;
   e)  Roadway Construction;
   f)  Dermal Exposure; and
   g)  Manufactured Soil  Conceptual Model.

   a)  Organizational Issues

   Dr. Mary Fox

       Page 3-1 to 3-2
       Section 3.1.1, 3.1.2 are repetitive of Chapter 2 - not needed in this chapter.

       Page 3-7

       Section 3.2 Benchmarks and criteria can also be re-located to appropriate sections of
       the analysis.

       The problem formulation chapter should reflect the framework shown in Figure 3-4.
       Rather than describe exposure pathways - this is the place to describe the screening
       modeling approach. Section 3.3.4 jumps the gun and includes results of screening
       analyses, listing constituents modeled as a result of screening.

   Dr. Donna Vorhees

       Provide graphical overview of the assessment. A single graphic that shows steps in
       COC selection, the deterministic screening analysis, the probabilistic analysis, and the
       relationship between the deterministic and probabilistic analyses would help improve
       clarity of the document. Figure 3-4 is a start, but is missing important information about
       how COCs were selected in a manner that ensures cumulative risks (i.e.,  risk across
       exposure pathways and chemicals that together might cause an adverse health
       outcome) associated with use of SFS do not exceed levels of concern. This
       understanding is important for states to determine whether use of SFS will meet their
       risk management goals.

       The assessment includes a "problem formulation" section, which discusses the purpose
       and scope of the assessment and defines the primary question and assessment
       endpoints to be addressed: "determine whether the proposed unencapsulated uses of
       SFS have the potential to cause adverse health or ecological effects (for this


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

       assessment, the Authors used the following risk management criteria: 10~5 risk for
       cancer effects and an HQ of 1 for noncancer and ecological effects)."

The Authors agree that certain portions of each chapter were repeated in the subsequent chapter.
However, that was intentional; the goal of these sections was to allow a reader to pick up at the
beginning of any chapter without having to go back. In other words, the chapter was meant to
stand alone for readers who did not have the time to read the whole document in one sitting.

With regards to the benchmarks and criteria, these data have been removed from Chapter 3,
Problem Formulation, and now are presented only in the relevant sections of the analysis.

 To better orient the reader, a figure has been added to Chapter 1 to depict the SFS assessment
framework and the fact that it is comprised of five key components: (1) SFS Characterization;
(2) Problem Formulation; (3) Analysis; (4) Risk Characterization; and (5) Conclusions. In
addition, the Analysis Plan (section 3.2) has been rewritten to clearly communicate the goal of
the analysis and provide an overview of the steps taken to accomplish this goal. The Authors
have also replaced Figure 3-4 with the figure shown below to better depict the relationship
between the different stages of analysis. The discussions in section 3.2 have also been revised to
provide an overview of the stages shown in the figure.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

| n
'•£ 1

"O "ff
8! 1
« o

re LJ



NO / Leac
<— — — < da
^\ avail

Constituents in SFS



hate^x ^
ta > o
able? / -c
/ ~°

r 1
\ I
/ Above \ o
^ NU / drinking water \ •
\ orde
^\ crite

rmal ^H
jon? //


NO / He
1^ 	 r bench

alth \,
mark ^H
^^avai able?/

1 /
NO / F
1^ / SCR
^x mode



EEN3 \


Soil Pathways

r/ Above \^
^detection limit?^^B




8) .E
5 «
X" "o
°- o
W ^ G
(0 •- sc
= a: (9
< A

model (IWEM)/
refined model

Probabilistic air


/^ \,
X^ Soil \,

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   Dr. Donna Vorhees

       / assume that this question relates to selection of chemicals of concern and not
       chemicals that should be measured in SFS. In general, the method seems reasonable,
       but is spread out across Sections 3, 4, and 5, which makes it slightly difficult to follow.
       On pages 1-2 and 1-3, the Authors indicate that "This report is intended to provide states
       with a sound scientific basis with which to evaluate the potential risks to human health
       and the environment associated with the beneficial use of SFS in soil-related
       applications." This goal could be achieved more easily by succinctly explaining the
       various screening steps that are described in Sections 3, 4, and 5 that resulted in
       elimination of chemicals from the list ofCOCs and, in some cases, elimination of entire
       exposure pathways.


As noted above, the Analysis Plan section 3.2 has been greatly enhanced with the addition of
new text and a figure that depicts the  different stages of the analysis, including Phase I:
Identifying constituents of concern. As discussed in the revised report, the Phase I Analysis was
used to identify COCs for three potential exposure pathways: (1) groundwater, (2) inhalation,
and (3) soil defined as both direct ingestion and indirect exposure via produce. Section 3.2 of the
report has been revised to discuss the steps taken to identify the COCs for each of these

   c)  Highly Exposed/Sensitive Subpopulations

   Dr. Donna Vorhees

       The problem formulation does not include discussion of highly exposed or highly
       susceptible populations, but this discussion appears later in Sections 5.3.2 in the context
       of the PRA (i.e., home gardeners more exposed than the general population) and 6.3.5
       (i.e., discussion  of the potential for plants to take up concentrations of cadmium and
       selenium that pose a concern for human health).

       Highly Exposed Populations.  The document would benefit from a clear discussion of the
       population targeted for evaluation - not just identifying the scenario (e.g., home
       gardener) - but the degree of exposure. The Authors explain that the analysis is intended
       to be protective  of the 90th percentile exposure level, but do not clearly define this
       criterion until Section 5.1.

       Susceptible Subpopulations.  The screening analysis does not explicitly evaluate
       childhood exposures, but the  probabilistic exposure assessment includes four age
       groups for individuals younger than 20 to account for variation in exposure over this
       time. The problem formulation appropriately discusses the importance of evaluating
       children separately from adults,  given their potentially higher intake to body-weight
       ratios. These groups do not include the 0-1 year old life stage, and the Authors assume
       that this exclusion overestimates risk. How is risk overestimated if the intent is to
       evaluate risk to an individual who is more than 1 year old? If the intent is to evaluate
       someone from birth, then why not do so? At a minimum, the significance of ignoring this
       age group should be discussed (i.e., explain that there is no exposure from birth to  about
       six months, when babies typically do not eat solid food, and review what is known about
       body-weight normalized ingestion rates for babies 6 months to 1-year- old).


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
Home gardeners who get fruits and vegetables from soils manufactured with foundry sands are
already more exposed than the typical population. In addition, the Authors evaluated risks to
children of home gardeners. Since home gardeners and children tend to have higher exposures,
the children of home gardeners are likely to be a relatively highly exposed subpopulation, and
thus protective of many other highly exposed subpopulations. The degree of this exposure can be
seen when compared to that of the general population. The original discussion in Section 3.1.5
has been modified as follows:
          "In addition to these overarching assumptions, the risk assessment of
       unencapsulated SFS uses was predicated on a number of conservative
       assumptions intended to ensure that the results could be used to support
       management decisions with a high degree of confidence. That is, the assessment
       was intentionally designed not to underestimate the potential risks to human
       health and the environment.

       •  The exposure scenarios focus on sensitive populations with respect to
          behaviors that tend to increase exposures. For example, the home gardener
          scenario represents adults and children that will have a relatively high level of
          direct contact (e.g., incidental soil ingestion) and indirect contact (e.g.,
          ingestion of home grown produce) when compared to other populations

       •  For carcinogenic (i.e., cancer-causing) constituents, the target cancer risk was
          defined as an excess lifetime cancer risk of 1 chance in 100,000 (i.e., IE-OS).
          For constituents that cause noncancer health effects, the target hazard level
          was defined as a ratio of predicted intake levels to safe intake levels—the
          HQ—of 1

       •  The Phase II modeling (explained further in Section 3.2.2, below) used the
          upper end of the exposure concentration distribution (i.e., groundwater
          screening modeling used  the 90th percentile receptor well concentration, and
          refined surface and groundwater modeling used the 90th percentile of the
          exposure distribution) rather than a central tendency measure

       •  Exposure assumptions used in the risk modeling were designed to likely
          overestimate, rather than  underestimate, potential exposures. For example, the
          exposure estimates from ingestion of home-grown produce assumed that the
          receptor consumes a very large amount of produce because the total produce
          diet is the sum of multiple produce categories (e.g., root vegetables, leafy
          greens). This implies that (1) all of these categories can be grown in the 0.1
          acre garden in the same season, (2) all of these categories are consumed at
          relatively high rates, and  (3) all these categories are consumed year round

       •  For effects to ecological receptors (e.g., plants, animals, soil invertebrates),
          conservative environmental quality criteria (i.e. EcoSSLs - see section 4.4.3
          for more on the conservative nature of these screening levels) were used in
          defining the target hazard levels

       •  The home garden was accessible to all residents, including children at all
          times; and


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       •  The addition of SFS manufactured soil (containing SFS at 50% of the soil dry
          weight) to the home garden essentially replaced the existing top 20-cm layer
          of local soil."

The Authors also note that the screening levels used in the assessment were developed by EPA
using the more protective  adult and child exposure factors. The assessment has been modified to
reflect this point. Thus, the screening analysis does take more sensitive subpopulations (children)
into account.

EPA has invested considerably in the development of distributions of the 1997 EFH,
subsequently updated in the 2011 EFH. The 2011 EFH currently represents the best available
consumption data. The CSEFH was published in September of 2008, and the data presented
therein generally agree with the 1997 and 2011 EFH. The exposure parameters used in this
assessment have been updated to reflect the CSEFH and 2011 EFH data.

The Authors acknowledge that evaluating a child starting in year 1 would not overestimate risk
for children of that age group. However, applying the constant starting year of exposure for the
child receptor does maintain the conservative nature of the assessment, given that exposure could
start at any point during childhood (i.e., from 1 year through year 19). The purpose of
implementing a start year  of 1 is to capture and bracket the risks for two distinct receptor age
groups. Evaluating children separately from adults is important given their potentially higher
intake to body-weight ratios.  Allowing the start age to vary for the child introduces the
possibility of the childhood exposures overlapping the adult exposures. For example, if the
child's starting year of exposure is at age 15 and the exposure duration is 8  years, the exposure
would span both childhood and adulthood. Thus, the clear distinction between the child and adult
exposures is not maintained.

With regard to infant exposure, EPA guidance is available to assess exposure via the breastmilk
pathway. The COCs for the soil/produce pathway include nickel, manganese, lead, and arsenic.
Current studies have not shown any of these constituents to be of significant concern via the
breastmilk pathway.  When discussing the assumptions built into the Conceptual Model for
probabilistic modeling, the Authors have modified the following text in Section to
improve clarity  and to address the absence of infant exposure modeling:
       "The adult was 20  years old when exposure began, and the child was 1 year of
       age when exposure began. Application of these start ages maintains the
       conservative  nature of this screening assessment. Infant exposure (i.e., 0 to  1 year
       of age) via the breastmilk pathway was not evaluated under this modeling
       scenario given that none of the metals included in the probabilistic modeling
       phase have been identified in current studies as being of significant  concern via
       the breastmilk pathway."

   d) Storage Pile Conceptual Model

   Dr. Ken Barbarick
       An oversight or weakness of the conceptual model is not considering leaching of the
       components  studied from the "storage pile". This process may not pose a risk; however,


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       it should be discussed including documentation that is available. A series of column
       "batch" leaching studies could be utilized to determine the extent to which any
       constituents are transported. Breakthrough curves could help estimate how many pore
       volumes of water would need to move through simulated piles to move significant
       quantities of each component. These data then could be included in the pathway

   Dr. Mary Fox
       Regarding Figure  3-2, The Blending  Site Model. There is a footnote to the figure
       explaining that deposition of particles and subsequent contact and ingestion were not
       quantified because it was assumed that related exposures would be insignificant
       compared to the manufactured soil (home garden) model. I agree with this for the human
       receptor, but the justification may not hold for wildlife.

With respect to Dr. Barbarick' s comment, leaching from a storage pile of SFS represents a much
lower risk to groundwater than the home garden because, by definition, the temporary storage
pile would remain in place for a relatively short period of time before use. Moreover, it is
reasonable to assume that the storage piles would be covered to some degree to prevent the pile
from becoming saturated with water (something that would greatly increase the weight of the
pile and make transport more difficult). Importantly, because this material has economic value, it
is expected that the facility would transport the SFS as rapidly as possible to generate revenue. In
addition to these considerations, Section 2.5.4 shows that many constituents were either not
detected in leachate sampling, or were detected at levels below health screening criteria.

Some constituents were detected in leachate  above health screening criteria, indicating that
drinking the leachate directly could result in  adverse health effects. For these constituents, the
Authors believe that the home-gardener scenario represents a much greater potential risk to
groundwater because (1) the SFS would remain indefinitely in the garden,  (2) the SFS is
incorporated into the  soil rather than sitting on top of the soil, (3) the garden presents a much
larger footprint (approximately 405 m2) than the temporary storage pile (assumed to be 150 m2 in
size), and (4) the underlying soil in a garden  would be expected to have a higher hydraulic
conductivity associated with agricultural soils (versus a compacted soil or concrete pad used for
the temporary storage of SFS). Also, the Authors have modified the text in Section 3.1.5 as
follows to provide additional support for the  assertion that the  potential for groundwater
contamination associated  with the use of SFS in manufactured soil by the home gardener
represents a much greater potential risk:

       •   "The home garden scenario is likely a much greater risk to groundwater than the other
          scenarios because (1) the SFS would remain in the garden indefinitely, (2) the SFS is
          incorporated into the soil rather than sitting on top of the soil, (3) the garden presents
          a much larger footprint (approximately 405 m2) than the temporary storage pile
          (assumed to be 150 m2 in size), and (4) the soil underlying a garden would likely
          have a higher hydraulic conductivity than a compacted soil or concrete pad used for
          the temporary  storage  of SFS.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications

       •  Because SFS and manufactured soils have economic value1, blending sites would
          process the SFS as rapidly as possible to generate revenue. This means that (1) the
          temporary storage pile would remain in place for a relatively short period of time
          before soil blending, and (2) the storage pile would likely be managed to protect the
          material's value and workability (e.g., use of a temporary cover to prevent loss due to
          runoff, and prevent the pile from becoming saturated with water).

       •  Commercial blending facilities demonstrate the greatest potential for nearby human
          inhalation exposures, because they tend to work with larger volumes of feedstock and
          product (thereby emitting greater volumes of particulates) and conduct operations
          throughout the year.

       •  The economics of purchasing, transporting, and applying SFS-manufactured soil
          would make its large-scale agronomic application untenable - farmers could not
          afford it.2 Other potential agronomic uses for SFS (e.g.,  to improve soil texture)
          involve application rates that would result in SFS concentrations lower than the
          assumed 1:1 blend (i.e., the soil is 50% SFS, by weight) in SFS-manufactured soil."

With regard to the commenters' concern over wildlife exposure, a storage pile of SFS represents
a much lower risk when compared to the home gardener scenario  due to the transitory nature of
the pile as discussed above and in the revised risk assessment report. Furthermore, the soil
concentrations (50% SFS) in the home gardener scenario should adequately screen for deposition
to adjacent soils because it would be unrealistic to think that these emissions would ultimately
lead to a soil mix of 50-50 off site.

In addition, the Authors note that the sands are below background levels for many of the trace
elements. This reduces the likelihood that sensitive ecological receptors would experience
significant effects.

   e) Roadway Construction

   Dr. Mary Fox
       With the following exception, the conceptual models capture the relevant sources,
       pathways and receptors: Figure 3.1 should include roadway construction or construction
       operations (i.e., moving SFS from storage to road building area) as a source with
       dispersion in air and deposition to soil as pathways.

       I am not sure that the assumptions regarding engineering controls on the storage pile for
       the roadway subbase model are reasonable.  Therefore, particulate emissions and runoff
       should be considered for evaluation. However, the  nature of roadway construction is
       likely temporary or intermittent, which would reduce concern about this source and
       related pathways. The temporary nature of construction activities is discussed in Chapter
       6, but it should be included in Chapter 3 along with the  descriptions of the conceptual
1 In 2007 manufactured soil sold for approximately $21.50 yd"3 (cost of product and delivery), or about $22,800 A"1
 for a 20 cm-deep layer (Kurtz Bros., Inc. 2007).
2 See previous footnote.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
The Authors agree with commenter that exposures from roadway construction using SFS will
typically be temporary in nature, and thus would not expose receptors to the chronic levels of
toxins necessary to induce health effects. In response, the Authors have updated Section 3.1.5 of
the original risk assessment to state the following:

       •  "For the temporary storage and use of SFS, indirect exposure pathways (e.g.,
          air emissions to soil deposition to soil-to-plant uptake to ingestion) would be
          unlikely to produce significant exposures because

          -  there would likely be engineered controls to prevent the loss of valued
             commodities, such as SFS feedstocks or blended soils,
          -  few chemical constituents have been shown to biomagnify in terrestrial food
          -  the time to reach steady state with respect to plant and animal concentrations
             would be insufficient, so bioaccumulation would be limited, and
          -  releases during roadway construction using SFS would be temporary and
             intermittent and, as a result, the potential for exposure will be very limited."

   f)  Dermal Exposure

   Dr. Donna Vorhees
       On Page 3-6, the Authors explain that "Dermal contact for the groundwater and soil
       pathways was excluded because available data indicate that the contribution of dermal
       exposure to soils to overall risk is typically small" based on results of a risk assessment
       conducted 14-15 years ago that reportedly involves only exposure to soil. This is not
       sufficient justification for excluding the soil dermal and groundwater dermal exposure
       pathways from further analysis. Did the cited risk analyses include the same exposure
       pathways and quantitative assumptions,  i.e., are they directly relevant to the current
       assessment? I doubt that the large differences between dermal exposure and other
       pathways cited in Note #2 on this page apply to a COC such as arsenic in the context of
       this assessment. Such short cuts might be technically justifiable in some instances, but
       the goal of this assessment is to provide states with the risk information they need to
       reach decisions. The best way to do this is to quantify risk from all exposure pathways or
       to quantitatively demonstrate within the document (not by reference to an older risk
       assessment with no explanation of its relevance) that the pathway does not warrant
       further analysis.

The Authors agree that when  SFSs are used in manufactured soils for home gardens, the
potential exists for dermal contact with soils and groundwater contaminated via leachate. For this
reason, the Authors have performed additional screening assessments to evaluate these potential
exposures for the COCs identified for the groundwater (Section 4.2.1) and soil  pathways
(Section 4.4.3).  The Authors have added new text to the report describing the methodologies
used to evaluate dermal exposures to contaminated soils and groundwater, and presenting the
results. As discussed in these additions, the soil and water dermal exposures were found to be
well  below a level of concern for all of the evaluated COCs.


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

Additionally, the Authors have updated the conceptual models depicted in Figures 3-1 through
3-3 to include dermal exposures. The new Analysis Plan (Figure 3-4) also includes dermal

   g) Manufactured Soil Conceptual Model

   Dr. Ken Barbarick

       The "Manufactured Soil" conceptual model is a highly conservative approach that will be
       protective of the other conceptual models. A 20-cm deep soil mix with 50% spent
       foundry sand is highly unlikely. The material and incorporation of a 50% mix to this depth
       would be expensive.

       The "Home Gardener" scenario is the best choice for modeling since it would pose the
       greatest risk to an individual. The comment in #7 [see comment in 5.e below] regarding
              should be considered.
       The "Home Gardener" scenario is very conservative; it almost represents a worst-case
       scenario. The assumptions for the general population consumption and independence of
       the ingestion pathways are appropriate.

   Dr. Mary Fox

       / agree that the Manufactured Soil conceptual model can be considered protective of the
       other conceptual models for human receptors. See above comment about the blending
       site model and exposures to wildlife from deposition of particles leading to ingestion.

   Dr. Charles Harvey

       Yes, [the Manufactured Soil conceptual model] appears "protective. "

       The major sources, pathways,  and receptors are included.

       Page 1-4, paragraph 3

       What does the spatial scale of the risk assessment mean? The size of the garden plots?
       The extent to which SFS is applied over a geographic area?

       Page 3-7, paragraph 1

       The statement that SFS will not be used for agronomic purposes is not convincing. It
       may be true that economics will always limit the use of SFS by farms, but I see no
       concrete evidence to support this contention. How expensive is SFS, and how much is it
       worth to improve soils for a farm? If SFS were used for a farm, then clearly a larger plot
       or garden area would need to be considered for the groundwater and home gardener

   Dr. Donna Vorhees

       The conceptual models are plausible and appear to include relevant sources, pathways,
       and receptors. However,  the Authors should document the sources of information (e.g.,
       stockpile management practices) used to construct conceptual models. Why do the
       Authors assume that engineering controls will prevent runoff but not fugitive dust? Also,
       engineering controls are not likely to be used for home gardens as appears to be


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       assumed in Figure 3-3. Where are direct links between SFS-containing materials (e.g.,
       manufactured soil on a garden/field") and receptors (e.g., home gardener) that ingest or
       come into dermal contact with it [Note: This direct contact pathway is shown correctly
       later in Figure 5-3]? Finally, the conceptual models are somewhat confusing, e.g., the
       legend suggests that dashed lines are used only for the surface water runoff pathway,
       but dashed lines are used for other pathways.

       Home gardeners are assumed to be exposed via the following exposure pathways:

       1.  inhalation of SFS emitted from soil blending operations
       2.  ingestion of groundwater contaminated by the leaching of SFS constituents
       3.  incidental ingestion of manufactured soil
       4.  ingestion of fruits and vegetables grown in the manufactured soil

       Except for the concern expressed below about exclusion of the dermal contact with soil
       and groundwater pathways, these exposure pathways are appropriate. There is also the
       potential for home-produced poultry, dairy, and beef. Did the Authors consider this
       possibility (e.g., links to several newspaper articles regarding the increase prevalence of
       backyard chickens can be found at: http://www. backvardc hie kens. com/LC-links. html),
       assuming that there is any reason to use manufactured soil for grazing areas?

       The Authors focus on SFS use in manufactured soil applied to gardens because this
       application is expected to result in the highest exposure.  Therefore, if exposure to
       manufactured soil is not associated with significant risk, then other applications also will
       not be problematic. If the  description of possible uses for SFS-containing materials is
       accurate, then the conceptual model for manufactured soil use appears to be protective
       of other SFS uses. However, it would  be useful to include a section that describes
       current controls and possible future controls on SFS use, if any, to support this
       assumption. How much SFS is produced annually? How much might end up in
       manufactured soil and what fraction of agricultural land used for food production  might
       ultimately have SFS-containing manufactured soil placed on it? The answers to these
       questions are relevant to  the assumption that the home garden pathway represents  an
       upper bound of possible exposure. Could a home gardener also be exposed from what
       they buy in the supermarket and/or the local community supported agriculture farm?

       Independence of Ingestion Pathways. "Sub-pathways include the incidental ingestion of
       soil, as well as the ingestion of exposed fruits (e.g., strawberries),  protected fruits (e.g.,
       oranges), exposed vegetables (e.g., lettuce), protected vegetables (e.g., corn), and root
       vegetables (e.g., carrots)" (page ES-4). On page 6-31, the Authors argue that "it  would
       be unlikely that a person would consume a high-end amount of root vegetables and leafy
       greens and apples that were all grown from the same garden." This statement might be
       true, but the Authors do not provide any data that substantiates this assumption.  Instead,
       they refer generally to consumption rates being high.


The Authors acknowledge that Dr. Barbarick, Dr. Fox, and Dr. Harvey find the home gardener
conceptual model to be protective. No further response is necessary.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
To address the potential for confusion pointed out by Dr. Harvey as it relates to the spatial scale
of the risk assessment, the Authors have clarified the question in Section 1.1 (and Section 6.2) as
       "Will the addition of SFS to soil result in an increase in the metal concentrations
       in soil relative to background levels, and how should the results of the risk
       assessment be interpreted across varied national soils?"

Regarding Dr. Vorhees questions regarding SFS production and potential use volumes, the 2002
Economic Census estimates that approximately 2,500 foundries operate nationwide. While this
number is likely to have declined due to recent economic trends, it is likely still close to the
actual number of operating foundries. In U.S. EPA (2008), it was found that the "metal-casting
process generates approximately 9.4 million tons of foundry sand annually."

However, this amount would include non-silica foundry sands and foundry sands generated by
brass and bronze foundries, etc. In addition, U.S. EPA (2008a) found that SFSs could be used in
higher price markets (e.g., fill, concrete),  could be located too far from  a soil blender to be
economically feasible, could be more expensive than competing native soils (e.g., parts of North
Carolina) etc. Therefore, the actual amount of SFS available for use in manufactured soil would
be much lower. With respect to agricultural use, the subject paragraph has been modified to
clarify and further justify  the position (see the final bullet in the Response to 4d, above).

Dr. Vorhees points out that the report should document the sources of information (e.g., stockpile
management practices) used to construct  conceptual models. For example, she asks why
engineering controls will prevent runoff, but not fugitive dust  The text inset below has been
added to the roadway subbase discussion as a footnote to support that runoff controls are a legal
requirement and that some of the same management practices will  also control fugitive dust as
required by the Clean Air Act. In the case of the blending site, fugitive  dust emissions were
considered as a release mechanism because the blending processes themselves, rather than
storage conditions, generate the emissions. Blending operations would also be occurring on a
much larger scale and, thus, would pose a higher risk than under the roadway subbase scenario.
       (footnote text) "Runoff controls are a legal requirement under the National
       Pollutant Discharge Elimination System (NPDES) that is part of the Clean Water
       Act. Most states have been authorized to implement the NPDES stormwater
       program (http://cfpub.epa.gov/npdes/stormwater/authorizationstatus.cfm),
       although some areas (e.g., tribal lands) remain under the direction of EPA. The
       NPDES regulations establish best management practices (BMPs) for any source
       of sediment, from sites or operations (e.g., construction, agricultural, or
       industrial), that might impact surface waters. Many of the BMPs applicable to the
       control of runoff are similarly used to control fugitive dust emissions as required
       under the Clean Air Act."

With regard to Dr. Vorhees comment concerning the use of engineering controls in a home
garden, the footnote on Figure 3-3 has been modified, as shown in  the text below, to clarify the
assumption that controls would be imposed to protect the gardener's investment in manufactured

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       (footnote text) "The scenario assumes that the home gardener would impose
       controls to prevent significant runoff/erosion of manufactured soil from the

The Authors have updated the conceptual model figures and footnotes in Section 3 to better
communicate the scenarios.

The Authors also acknowledge that they did not consider the possibility of beef, diary, or
backyard chickens in the home gardener scenario. As discussed above, the use of SFS in an
agricultural setting is limited due to economic feasibility. Because of the soil-plant barrier,  the
low potential for uptake of metals that are largely unavailable, and the relatively limited amount
of animal products that could be raised on soils amended with SFS, the Authors do not believe
that this represents a significant limitation of the analysis.

To address the comment regarding the clarity of the statement "it would be unlikely that a person
would consume a high-end amount of root vegetables and leafy greens and apples that were all
grown from the same garden," the Authors have revised the statement as follows:
       "It would be unlikely that a person would consume a high-end amount of root vegetables
       and leafy greens and apples that were all grown from the same garden because (1) all
       types of produce cannot be grown in the same season, (2) there are regional
       characteristics (e.g., soil type, precipitation) that strongly influence what types of crops
       can be grown, and (3) there are agronomic limits as to how much produce can be grown,
       harvested,  and consumed that are not reflected in the exposure factor data."

The Authors believe that this revision clarifies the above statement; however, it should be noted
that the Authors used data from EPA's Exposure Factors Handbook and, therefore, are confident
that the data are appropriate for the intended purpose. We believe that the paragraph and other
changes to the report make it clear that, taken together, the consumption rates for fresh produce
are conservative (tend to overestimate the actual consumption rates) and appropriate for the
purposes of developing a conservative risk assessment screen for the produce ingestion pathway.
Further, we explored this conservatism by running the model for alternative scenarios (e.g.,
general population) to provide quantitative insight into the risk estimates for receptors that
represent more typical  consumption rates for fresh, home-grown produce.

5) Screening and  Modeling

The peer-review comments relating to screening of constituents and pathways are grouped  into
the following subcategories:
   a)  Positive Comments;
   b)  Soil-Blending Site Distances;
   c)  Groundwater Model;
   d)  Consumption Model Protectiveness;
   e)  Constituents Contributed by Soil;
   f)  Unitized Exposure Estimates; and
   g)  Transparency Issues.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications

   a)  Positive Comments

   Dr. Ken Barbarick

       Yes, the screening steps were appropriately conservative. Model equations are based
       on documented modeling research. The elimination of the TCLP test for "Ingestion of
       Groundwater" pathway is appropriate. The study also provides good justification for
       which metals were retained to determine risk of exposure.

       The study used different screening levels developed at Oak Ridge National Laboratory.
       The risk assessment execution is solid. The screening levels from Oak Ridge National
       Laboratory are commonly used and are the best information available. They are
       sufficiently protective for the risk assessment used in this study.

   Dr. Mary Fox

       For the most part, the deterministic screening modeling was straightforward and clearly

       The air and groundwater screening steps were clearly designed to be conservative, e.g.,
       95* %ile sampling data were used for modeling and comparisons. Selection of
       constituents to evaluate in drinking water scenario is conservative. Contaminants were
       retained because LOD for leachate testing falls above the screening reference levels.

       The screening of soil and produce ingestion pathways was trickier because it involved
       the "dilution" of SFS concentrations due to mixing with other soil components in the
       manufactured soil and consideration of multiple sub-pathways.

       To address the issue of multiple sub-pathways of exposure the Authors divided the SSL
       health screening benchmarks by 10 to derive an adjusted SSL that allows for multiple
       pathways of exposure.  This is an appropriate and conservative approach.

   Dr. Donna Vorhees

       The screening assessment is based almost entirely on the conceptual model for
       manufactured soil use on a home garden because this use is assumed to be associated
       with the highest degree of exposure. The exception is the use of a soil-blending
       operation to represent an upper-bound exposure estimate for the inhalation of fugitive
       dust pathway.

       The assessment appropriately includes deterministic methods. The extensive use of
       screening in lieu of "forward" risk calculations might make risk communication a


The Authors acknowledge the above comments. The use of initial screening in lieu of "forward"
risk calculations is a standard, accepted practice in environmental health risk assessment, and the
Authors do not anticipate risk communications challenges. No further response is necessary.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   b) Soil-Blending Site Distances

   Dr. Charles Harvey
       The calculated risks from inhalation were based on a minimum distance of 500 m
       between the nearest residence and the source. This choice of value for the downwind
       distance does not appear to be conservative, especially relative to other selected
       parameter values. The choice is based on a single areal photograph of a blending
       facility. It is reasonable to suspect that, if more sites were considered,  some would have
       closer distances to the nearest residence.  For a conservative screening calculation, the
       assessment should use a minimum distance closer to 100 m. For the groundwater
       model, the choice ofalm distance from a garden to a drinking water well appears to
       have been an attempt to be conservative (however, see comment above). The same
       philosophy was not used for the choice of the distance between blending site and the
       nearest house. A distance of 100 m seems like a reasonable, conservative choice.

       Decreasing the assumed distance to the closest residence may push the 95th percentile
       for manganese over the screening concentration. At 500 m, the calculated value of 501
       mg/kg is only a factor of two less than the screening concentration (Table 4.4). Such an
       outcome would complicate the overall assessment. However, it could be very useful for
       devising future regulations for building  blending facilities.

       Issues related to manganese poisoning from inhalation have been considered in studies
       on the dangers of mining dust.

Dr. Harvey suggests that the assessment should use a minimum receptor distance closer to 100 m
instead of the 500 m distance applied in the assessment. However, we believe that the distance of
500 m is appropriate for screening purposes. The goal of the assessment was to model  a
reasonable maximum exposure (RME) scenario. In defining this scenario, the Authors identified
or developed parameter values that were consistent with high-end emission and dispersion
conditions, but not "worst-case" conditions. As shown Table 2, several aspects of the modeling
approach maintained the high-end, conservative nature of the assessment. For instance, modeling
was performed using EPA's recommended conservative screening model  SCREENS. The full
range of meteorological conditions and wind directions were  examined to ensure that modeling
options identified the highest concentrations. This model generated short-term, maximum 1-hour
air concentrations. These short-term concentrations were then combined with chronic health
benchmarks to develop conservative screening levels. Lastly, these screening levels were
compared to the 95th percentile SFS concentrations to ensure that the concentrations did not pose
an unacceptable risk to human health. The Authors believe that  compounding these high-end
modeling elements with a receptor distance of 100 m would result in an unreasonable "worst-
case" scenario and not an RME scenario. The Authors would also like to point out that dividing
all current inhalation screening concentrations  by a factor of 5 (i.e., to reflect a reduction of
receptor distance from 500 m to 100 m) would not change the conclusions of the analysis.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
             Table 2: Parameters for Screening Level Inhalation Assessment
Model Selection
Emission rates (g s"1 m~2)
Height of storage pile (m)
Length of storage pile (m)
Width of storage pile (m)
Receptor height (m)
Urban or rural
Search for maximum
Choice of meteorology
Distance (m)
Health Endpoints
SFS concentration levels
Calculated based on
SCREENS is the screening version of the
Industrial Source Complex model, version 3
(ISC3), used to calculate short-term,
maximum 1-hour air concentrations
Calculated using high-end wind speeds and
rainfall assumptions
Based on aerial photography
Based on aerial photography
Based on aerial photography
Representative of breathing zone for child
receptor (i.e., ground level)
Rural option selected based on observed
surrounding land use
Examining all directions ensures that the
maximum concentrations will be located
Under this option, SCREENS examines a
range of stability classes and wind speeds to
identify the "worst case" meteorological
conditions, i.e., the combination of wind
speed and stability that results in the
maximum ground level concentrations
Distance to the nearest resident based on
aerial photography
Benchmarks used to calculate the screening
level are based on the worst-case exposure
duration, and frequency of 24 h d"1, 365 d
yr1 were used for comparison to short-term
maximum air concentrations
95th percentile concentrations of
constituents in SFS were for comparison to
risk-based screening levels
   c)  Groundwater Model

   Dr. Mary Fox

       Numerous subsurface parameters for groundwater modeling were set to model default
       values. This is outside my area of expertise, but I wonder how these defaults influence
       the "national representativeness" of the groundwater ingestion pathway.

   Dr. Charles Harvey

       The groundwater modeling resulted in a result of "zero" for all estimated 90th percentile
       exposures (Table 5.1). First, "zero" fora chemical concentration appears a little peculiar
       - "zero" really means that the modeled results are at or below some minimum value that
       the model can accurately produce.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       The more important question about these findings is: why are the values so low? Why
       does the simulated leachate not reach the well, or why is it so greatly diluted? Some
       simple calculations are useful for approaching these questions. First, it is useful to
       consider how long it will take the leachate to reach the well. If we only consider the time
       to flow through the saturated zone to the well,  then the model parameters imply that it
       will take about a year for a conservative solute to reach the well from the  far upstream
       side of the plot. (In all model runs, the plot was 1 acre (-40 m x -40 m), the gradient was
       0.0057, and the hydraulic conductivity was  1890 m/yr. So, assuming a porosity of 0.25, T
       = (40 x 0.25)7(1890 x 0.0057) = -1 year). Thus, the modeled time for one of the solutes
       (e.g., arsenic) to reach the well will be longer, and perhaps much longer, than a year
       because the  model includes the transport time through the unsaturated zone,  and
       solutes are subject to sorption as parameterized by retardation factors. But, what time
       duration was modeled? The description states that the "land application unit was
       operated for  1 year," but for how long was the  leachate input simulated, and for what
       time period was groundwater transport simulated?

       Furthermore, what was the screened interval of the well? If concentrations at the bottom
       of the well were considered, then they would be "zero" because the bottom of the well is
       on a stream line that extends upgradient to a recharge source beyond the plot. For
       qroundwater concentrations below some depth in the aquifer, putting the  well  very close
       to garden plot is, in fact, not conservative -contaminants from the SFS will pass above
       the depth of the well because the well is so close to the garden plot. (For a stream line to
       extend from the plot to the bottom of the well, recharge would have to be  greater than
       2.5 m. Fora porosity of 0.25 again, and approximating stream lines as parallel, the
       recharge rate that will reach the bottom of the  10m aquifer in one year is, 10 x 0.25 =
       2.5 m. None  of the realizations should have such a large recharge rate, and hence,
       solute should not reach the bottom of the well in any of the realizations.) If only top levels
       of the aquifer are considered, then concentrations will rise more quickly after creation of
       the garden plot because leachate will reach the well quickly near the top of the aquifer.
       The "protective" approach would be to use the maximum concentration with depth.

       In summary,  there simply isn't enough explanation  of the model to understand whether
       the "zero" concentrations are a robust finding,  or whether they result from a peculiarity of
       the model setup.  This report does not make a convincing case that the groundwater
       modeling has been carefully considered. For example, hydraulic conductivity is the
       largest source of uncertainty in most groundwater models, yet in this probabilistic
       assessment that parameter is set to a constant value. In fact, it appears that this
       assessment would be better served by employing a simpler approach— that a
       sophisticated groundwater model may be unnecessary. Simple approximations of pore-
       water velocities and retardation factors would produce equally valid outcomes, and such
       an approach would be more transparent.

       Page 5-4,  last bullet point

       Why were the concentrations of antimony, beryllium, cadmium, and lead modeled at half
       their detection limits? The detection limit would be the appropriate "protective" value.


The Authors used fixed default values for the seven key parameters listed in Section 5.2.1 of the
original risk assessment. These model defaults, as well as the default input distributions, are from
national distributions, as explained in U.S. EPA (2002a). These parameters were peer reviewed


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
along with the model, as discussed in U.S. EPA (2002a). Thus, the Authors believe that these
default values are appropriate and representative for the nationwide, probabilistic risk assessment
conducted here.

In response to Dr. Harvey's questions about the modeling results, the model returns zero when
the modeled concentrations are less than 10~20 mg L"1. Footnote 'a' of Table 5-1 has been revised
to clarify this threshold. IWEM ran the scenario 10,000 times, and each iteration was modeled
for 10,000 years or once peak receptor well concentrations were identified, whichever came first.
The 10,000 model iterations would randomly choose different well screen depths within the
shallow aquifer. The aquifer was located between depths of 5.18 m and 15.28m below the
ground surface.

The Authors agree with Dr. Harvey that the combination of short distance to receptor well and
randomly chosen well depth (i.e. "screen interval") may not lead to conservative results. To test
the effects of this combination on exposure concentrations, additional probabilistic screening
modeling was performed at well distances of 15m, 30m, 50m, 75m and 100m. Numerous
changes have been made to Section 5.2 (Screening Probabilistic Modeling of the Groundwater
Ingestion Pathway) of the report to reflect the rationale behind the additional modeling, how it
was performed, and the results.

Also, a complete review was conducted of the probabilistic screening modeling of the
groundwater pathway. The review evaluated the use of IWEM as detailed in the report (e.g.,
choice of garden size, well distance, operational life, etc) to see if any changes would result in
more accurately representing the SFS home garden scenario. The review was designed to also
identify alternative peer-reviewed and publically available groundwater models, if any, that
could more accurately represent the SFS home garden scenario. As a result of the review IWEM
was retained as the groundwater model of choice, but several changes were made to how it was
used. Specifically, the size of the home garden was reduced to 0.1 acre to more closely reflect
home garden sizes, and the operational life of the garden (used by IWEM to track the duration of
metal's leaching from the garden) was increased to 40 years. Section 5.2 (Screening Probabilistic
Modeling of the Groundwater Ingestion Pathway) of the report has been modified to  reflect the
changes in modeling methodology and results.  A complete description of the review process and
its findings is found in Appendix A of this Response to Comments document.

The modified screening modeling found that arsenic exceeded the lowest available screening
level in the wet and central tendency climates. Arsenic was therefore retained for more refined
study. Complete descriptions of how the refined study was conducted, and the results, have been
added to the report in Section 5.3 (Refined Probabilistic Modeling of the Soil/Produce and
Groundwater Ingestion Pathways)  and Appendix J (EPACMTP Groundwater Modeling). The
results of both the probabilistic screening modeling and refined modeling have also been
included in constituent-specific subsections, as well as discussions of uncertainty, in Chapter 6
(Risk Characterization).

The Authors agree that it is useful to consider how long it would take for SFS manufactured soil
constituents leaching from the garden to reach the receptor well. If exposures via groundwater do
not occur in the same timeframe as exposures via surface pathways (e.g. ingestion of homegrown
produce), then it may be more  appropriate to assess potential health impacts separately.


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
Additional modeling was performed, including the use of retardation factors as Dr. Harvey
suggested, to determine if surface and groundwater pathway exposures would occur within the
same timeframe. The results of this modeling demonstrated that peak exposures from surface
pathways would not overlap with peak exposures via groundwater. References to this additional
modeling, or the implications of the results (i.e. not aggregating surface & groundwater
exposures), have been added to pages ES-5, ES-7, 3-13, 5-29, 6-13, and 7-3. The following
language was also added to Section  5.3.5:

       "An analysis was performed to evaluate anticipated arrival times to determine if the
       exposure through the soil ingestion pathway would overlap with exposure through the
       groundwater pathway. To determine the approximate timeframe when the peak
       groundwater exposure might occur, estimates were made of the time at which the
       contaminant plume would arrive at the receptor well and the time when the contaminant
       plume would finish passing the well. Arrival of peak concentrations would only occur
       somewhere within this time period. These estimates were based upon two additional
       outputs from the unsaturated zone transport simulation: 1) first arrival time of leachate at
       the water table and 2) cessation time of leachate arrival at the water table. Retardation
       effects were used to account for horizontal travel to the receptor well. The results of this
       analysis are summarized in Table 5-3.
                          Table 5-3. EPACMTP Arrival Times
                        of Plume at the Receptor Well for Arsenic
Arrival Time Zone (year)
       Based on the analysis, (see Appendix J for more details), it is unlikely that peak surface
       and peak groundwater exposures will occur within the same timeframe. For example, the
       earliest estimated timeframe for arrival of arsenic from the garden spanned from 29 to
       almost 400 years following the application of the SFS. It is therefore likely that the peak
       well concentrations will not occur until well past the timeframe for peak surface
       pathways exposures, and perhaps even past the timeframe of residency (i.e., exposure
       duration of the gardeners who originally applied the SFS manufactured soil). Therefore,
       separate screening levels were developed for the groundwater and soil pathways."

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
In addition, a detailed description of this additional modeling, and the results, has been added to
Appendix J (EPACMTP Groundwater Modeling) of the report.

With respect to the treatment of nondetects, the Authors assumed that there would be some
distribution of concentrations between zero and the detection limit. Because the midpoint of the
two extremes of such a distribution would be half the detection limit, the Authors determined
that replacing nondetects with half the detection limit would be appropriate. This also is
consistent with both U.S. EPA (1989) and U.S. EPA (1991). The Authors have modified the text
of Section 5.2.1 to state the following:

       "For arsenic, the higher of the 95th percentile leachate concentrations found by
       either SPLP or the ASTM leachate methods (0.018 mg L"1) was modeled.
       Antimony, beryllium, cadmium, and lead were not detected in any samples, and
       were therefore modeled at half their detection limits in accordance with U.S. EPA
       (1991b). Thus, their modeled leachate values were 0.02, 0.01,  0.005, and 0.055
       mg L"1, respectively."

   d) Consumption Model Protectiveness

   Dr. Mary Fox

       / believe the Authors took reasonable steps to develop models to represent the range of
       site conditions in the continental United States. This included  using regional
       meteorological data, modeling multiple soil types and climate  conditions and defining
       SFS use feasibility zones.

       The home-gardener scenario as described is probably conservative, but not necessarily
       a "significant overestimation (page 6-31)." Independence ofingestion pathways is an
       appropriate assumption.

       The stated reason for modeling the  general population was concern that the home-
       gardener scenario was overly conservative. I do not share that view. However, it is
       useful to have a range of estimates to represent other populations with moderate

   Dr. Donna Vorhees

       The Authors assume that manufactured soil is 50% SFS and explain that a higher
       percentage would not be feasible because it would be cost-prohibitive for a home
       gardener (i.e., see note 5 on page ES-6, which indicates that  blends "are more likely to
       include 5-10% SFS" for this reason), and the manufactured soil would not have the
       characteristics needed to grow plants (See Dayton et al. manuscript, in review). Still,
       over time, manufactured soil could be used repeatedly in a single location, so it makes
       sense to consider the potential fora higher percentage contribution of SFS. The
       assumption that SFS comprises 50% of manufactured soil is not certain, but does seem
       to provide an upper bound given soil requirements for growing plants.


The Authors have replaced the phrase "significant overestimation" with the phrase "as an
overestimate" in the revised document.


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
The Authors also note Dr. Vorhees' finding that the 50% SFS assumption provides an upper
bound for manufactured soil. No further response is necessary.

   e)  Constituents  Contributed by Soil

   Dr. Ken Barbarick
       One suggestion is to include the soil contributions to ConcMs. No doubt the contribution
       would be small in most cases; however, including this information provides a more
       thorough analysis.

The Authors note that the  scope of this assessment was to look at incremental risks due to
exposures to SFS constituents in soil-related uses. Thus, the constituent concentrations
contributed by soil are outside the scope of this assessment.

   f)  Unitized Exposure Estimates

   Dr. Mary Fox
       The rationale and approach to the probabilistic modeling of the soil/produce ingestion
       pathway (Section 5.3) is not clear. Why were unitized exposure estimates preferable to
       health risk estimates? More background on development and uses of unitized exposure
       estimates is needed. Why was 1 mg/kg chosen as the assumed concentration?

       Methodology for developing unitized exposure estimates needs to be explained more
       thoroughly including a specific example and references to other EPA uses of unitized
       exposure or risk estimates. Why is this approach necessary or preferred? How does the
       assumed concentration of 1 ppm relate to actual manufactured soils or what would be

The Authors agree that the introductory discussion on the soil/produce ingestion pathway does
not provide sufficient background for the modeling and, in particular, some of the terminology  in
the beginning of Section 5.3 has caused confusion. The term "unitized" in this context refers to
the use of a fixed, initial concentration of a metal or metalloid constituent in SFS that was used
as input to the Monte Carlo modeling simulations. The Authors have revised Section 5.3 to
summarize the individual steps taken to develop the target soil concentrations from the unitized
risk distributions and have added a new section (Section 5.3.1) that provides a detailed
description of the probabilistic modeling framework and explains why fixed concentrations—
rather than sampled concentrations—were used as input to the model simulations. Application of
the unitized approach for this assessment was appropriate because the modeling system is linear
with respect to concentration and a "unitized concentration" of 1 mg/kg could be used to
calculate the allowable concentration of specific metal constituents in SFS (i.e., representing
minimal risk). Importantly, calculating allowable constituent  concentrations in SFS provides the
states with a frame of reference with which to address the variability of chemical concentrations
and characteristics of foundry sands. Thus, the "unitized"  concentration of 1 mg/kg was chosen
arbitrarily as the initial concentration with which to scale to an acceptable concentration in SFS,


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications

defined by EPA as a 90th percentile hazard for the soil/produce ingestion pathway below a value
of 1 (note that any other fixed concentration would serve the same purpose). In addition to these
changes, the Authors have added the following footnote to Section 5.3 to demonstrate other EPA
uses of this unitized approach for risk assessments.

       (footnote text) "Similar unitized approaches have been applied under previous
       U.S. EPA risk assessments. For example, the unitized approach was applied in the
       Risk-Based Mass Loading Limits for Solvents in Disposed Wipes and Laundry
       Sludges Managed in Municipal Landfills. This risk assessment and the unitized
       approach have been extensively reviewed internally and externally by the OMB
       and the final rule based on this risk assessment, Solvent-Contaminated Wipes,
       was publish July 31, 2013 (78 FR 46448-46485)."

   g) Transparency Issues

   Dr. Mary Fox

       Software used and specifications of the probabilistic modeling including number of
       iterations and type of sampling (Monte  Carlo or Latin Hypercube) should be provided in
       the main text or as an appendix. This information is needed for transparency. A
       complete evaluation of the probabilistic modeling cannot be conducted without this

       Joint probability approach for determining the combination of site conditions evaluated in
       the probabilistic modeling is not well described in Chapter 5 or Chapter 3. How is the
       joint probability approach implemented within the modeling framework?

   Dr. Charles Harvey

       Also, as a more specific comment, the report should better illustrate the Soil/Produce
       Ingestion Pathway model (Section 5.3). This model is an important part of the overall
       assessment and is bewilderingly complex. A simple way to bring some clarity to the
       model presentation is to illustrate a mass balance for the model. The flow chart of mass
       fluxes for the conceptual model is intricate, and as presented, it is impossible for the
       reader to determine the magnitude of the different fluxes. A mass balance for the model
       would illustrate  how much mass of a particular contaminant is applied,  and  then how
       much of this contaminant is transported through the different pathways. This would give
       the reader some notion of the importance of the different pathways. Also, constructing a
       mass balance is absolutely key to validating a model —the mass fluxes must sum to the
       mass loss. Thus, presentation of the mass balance would also provide some confidence
       in the workings of the model. This balance continues to hold when mean values across
       all realizations are used, and showing the average values may be the best way to
       illustrate the mass balance, although augmenting the averages with their standard
       deviations would improve the illustration.

   Dr. Donna Vorhees

       The assessment includes documentation of models, data, and assumptions used to
       perform all analyses, except as otherwise specified in responses to charge  questions.

       The assessment briefly describes EPA  models such that all work could be independently
       reproduced. Some additional discussion of the 3MRA model would be helpful for the


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       reader to understand the modeling in greater detail. However, it is possible for a reader
       to consult EPA guidance regarding the model, as well as the software itself.

       I could not find discussion of numerical stability of PRA model outputs. Does the 3MRA
       model provide any quality assurance output to check for such stability? If so, provide a
       summary in this assessment.

Dr. Harvey's request for a mass-balance equation and Dr. Fox's questions speak to the
probabilistic model documentation. As noted by Dr. Vorhees, further discussion of the models is
available in U.S. EPA (1999a, 1999b,  and 2003c). However, to improve transparency, the
Authors have revised Section 5.3 and added a new Section 5.3.1 to better communicate the
methodologies and modeling that was implemented within the probabilistic framework.
Additional language was also added to Section 5.3.4 to direct readers to a detailed description of
the refined model's mass balance structure in Appendix G.

The choice of 7,500 iterations in this analysis was based on the Authors' historical knowledge of
conducting Monte Carlo simulations using these models for EPA risk assessments. In response to
peer-review comments on stability, the Authors evaluated this assumption by performing a
stability assessment for the home gardener scenario. The results shown below for arsenic
demonstrate that performing 7,500 iterations adequately ensures the stability of the results at the
percentiles of interest (50th and 90th).

The Authors have included the arsenic example in Section 5.3.1, and have described the stability
test and present the results shown in the following table. The table shown in the figure presents
the absolute percent changes between  samples. As demonstrated by this figure, the model is
stable before 5,000 iterations for the mean, variance, and at the 50th and 90th percentiles.
Unitized Hazard Estimates
Total Ingestion: Child of Home Gardener
rd Estimates (unitless)
3 O
l tv» ui C
i 0.1 i
0.05 -
\ 	 5f 	

-^*— 7500

0 60 70 80 90 100
50th percentile
75th percentile
90th percentile
95th percentile
99th percentile
Percent Difference


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
6) Risk Characterization and  Uncertainties

The peer-review comments relating to characterization of the risks and their related uncertainties
are grouped into eleven subcategories:

   a) Findings/Conclusions;
   b) Soil Properties, Background, Phytotoxicity, and Soil Biota;
   c) Sensitivity Analysis;
   d) Variability versus Uncertainty;
   e) Inconsistencies with the Exposure Factors Handbook;
   f) Child-Specific Exposure Factors Handbook Not Used;
   g) Data Collection Uncertainty;
   h) Toxicity Value Uncertainty;
   i) Consumption Rate Uncertainty;
   j) Cumulative Risk; and
   k) Clarity.

   a) Findings/Conclusions

   Dr. Ken Barbarick

      The "Risk Characterization" section is very clear. The weight of evidence approach for
      (a) risk screening modeling and (b) uncertainties associated with state-of-the-science
      research provided the best assessment.

      The document does a thorough job of providing and interpreting information without
      hidden assumptions or preconceived notions. The risk assessment is "transparent."

      The assessment does support the report's conclusions that spent foundry sands can
      safely be used as an up to 50% manufactured or garden soil mix.

   Dr. Mary Fox

      / believe the assessment was conducted as the Authors report. The overall approach is
      reasonable.  The formulae appear correct and models (SCREENS, IWEM, 3MRA,
      ISCST3) used are appropriate. However, as detailed in answers to subsequent
      questions there are data limitations and issues with implementation of the probabilistic
      modeling that compromise the Authors' claims of conservatism and the assessment

      I cannot endorse the risk assessment findings and conclusions as presented in the peer-
      review draft. The probabilistic modeling analysis needs to be revised considering these
      comments and repeated.

   Dr. Charles Harvey

      The study also uses a "weight-of-evidence approach," and claims (p. 1-4) that it is
      "useful to consider exactly what this means," but does not appear to present a definition
      that clearly distinguishes this approach from simply conducting a good study. As best I
      can tell, the "weight-of-evidence" approach means a comprehensive study that brings

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       together all useful lines of evidence. But, I am left wondering if there might be something
       more to this phrase.

       The report makes a strong argument that SFS use is safe. However, the report would
       ultimately be more compelling, and certainly more useful, if it focused on providing the
       best description of the distribution of risks.

       Rather than present only 9Cjh percentile hazard estimates (e.g.,  Table 5.8), the
       assessment would benefit by presenting the entire histogram. Using histograms instead
       of point estimates has a number of advantages, as follows: (1) It would remind the
       reader that the estimates are the result of a Monte-Carlo simulation and give the reader
       a visual representation of the spread of the resultant distribution. (2) In the current
       presentation, there  is no indication of the skewness of the distribution - above the 90%
       cutoff, just how large are the values? As a hypothetical example, if the distribution is very
       skewed then more than half of the health risks could lie above the 90% cutoff,  and hence
       the approach taken in this assessment would miss the real danger. (3) Using the 90%
       cutoff is arbitrary. The full histogram offers the possibility of estimating other point

    Dr. Donna Vorhees

       Yes, the assessment supports the overarching conclusion that beneficial use of SFS can
       occur without significant risk to human health. However, the issues raised in response to
       other charge questions require attention.

       The assessment provides a clear discussion of how risk-based screening levels were
       developed, including a discussion of uncertainties that influence interpretation  of results.
       I also understand the utility of screening levels as opposed to "forward" risk calculations
       in this context where states and others might want to compare chemical concentrations
       associated with individual samples of SFS or SFS-containing materials to "acceptable"
       concentrations. However, as noted in response to other charge questions, this section
       could more succinctly address the general question of whether the assessment, in its
       entirety, ensures that cumulative risks are below levels of concern.

       The Authors conclude that "the results of the home gardener risk screening modeling
       should be considered as a significant overestimation of the actual  risks associated with
       SFS use." This conclusion might be true but is not substantiated adequately in the
       assessment as discussed in response to charge question #8.

       Overall, the execution of the risk assessment is adequate and excels in some respects.
       My responses to all charge questions highlight opportunities to improve the document.


The Authors acknowledge that several of the commenters found the conclusions to be accurate.
No further response is  necessary to these statements.

While Dr. Fox states her inability to agree with the conclusions due to specific issues, the
Authors have attempted to address all of these issues in the revised draft. Because most of these
issues were transparency issues, and the remaining reviewers agreed with the conclusions, the
Authors believe that the conclusions are scientifically supportable.


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
With respect to Dr. Harvey's first comment, the Authors acknowledge that the term "weight of
evidence" may not be the typical description of risk results, and thus, it has the potential to
confuse readers. Such terms have been modified to refer to "lines of evidence," which is both
accurate and more typical of risk assessment documentation.

The Authors do not believe that providing histograms of the unitized risk estimates per Dr.
Harvey's other request would be as useful. These distributions could not be converted into a
screening value for states to use and would be confusing to all but the most technical readers.
Thus, while taking a point estimate at the 90th percentile may be considered arbitrary by some, it
is a deliberately high-end value that allows the Authors to provide to the states a useful screening
value that is protective of human health and  the environment. Additionally, the Authors did
create a range of screening values by using three different consumption rates, which allows for
some flexibility among risk managers.

In regard to Dr. Vorhees comments regarding cumulative risk and the "significant
overestimation" language, these are addressed in 6.j below and 5.f above (respectively).

   b)  Soil Properties, Background, Phytotoxicity,  and Soil Biota

   Dr. Ken Barbarick

       The report presents a very thorough scrutiny of soil fertility, nonessential elements, and
       potentially toxic compounds.

       The soil background and phytotoxicity are adequate.  The impacts on general soil biota
       needs more detailed study.  For example, earthworms are mentioned as a group in terms
       of potential risks. Earthworms are a very diverse group of organisms who will more than
       likely respond differently to the potential risks associated with spent foundry sand
       additions to soil. This study probably did not have the resources to look at specific
       groups of biota, however. Shifts in microbial communities should be studied to determine
       if the "Home  Gardener" scenario encourages shifts between major microbial groupings
       such as bacteria and fungi and if particular individual  species of organisms are favored
       or harmed by the additions of the spent foundry sand. Good references for this approach
       are the following: Ritchie et al. (2000), and Schutter et al. (2001).

   Dr. Charles Harvey

       The assessment describes a broad and representative sampling of the research

       Finally, in several places, the report  emphasizes that background concentrations of
       metals in SFS do not appear to be much higher than  in natural soils, and therefore, the
       use of SFS poses no danger.  This may be true, but this statement should be tempered
       with several caveats. First, SFS could contain artificial organic contaminants left after
       heating the binding agents.  Second,  the metals could be in a less recalcitrant state than
       in natural soils.

   Dr. Donna Vorhees

       The comparison of SFS concentrations to USGS background concentrations in Figures
       6-1 through 6-4 is useful, although I suggest that axes on paired plots should be


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       consistent to facilitate the comparison. Treatment ofnondetect results should be
       specified on these plots and any other data manipulation that might influence the


Dr. Barbarick makes several good suggestions as to future research that could be conducted on
the topic of soil biota. However, the risk assessment was designed to use available data, and
conducting field studies on soil biota is well beyond the immediate scope of this study.

The Authors agree with Dr. Harvey that SFS could contain certain organic contaminants and
metals in a less recalcitrant state. For this reason, the study considers background concentrations,
the analytical data on potential releases, field and laboratory studies on ecotoxicology, screening
comparisons between SFS and health and environmental quality criteria, and probabilistic risk
modeling in developing the risk characterization. As suggested by Dr. Harvey, simply comparing
SFS concentrations with background concentrations would not provide adequate information to
support a risk characterization. However, the Authors believe that the lines-of-evidence approach
used to combine and consider all of these data supports the conclusions of this study.

The Authors understand Dr. Vorhees point about matching axes;  however, we believe that axes
as presented convey the information as intended. The purpose of these figures is to convey
information about the concentration distribution for each set of data and, therefore, we believe
that using axes that allow the shape of the distribution to be visualized is warranted.

As mentioned in 5.c above, the Authors have added discussion regarding the treatment of
nondetects to Section 5.2.1 of the revised document.

   c) Sensitivity Analysis

   Dr. Donna Vorhees

       The assessment incorporates appropriate sensitivity analyses and describes them


The Authors acknowledge the  comment. No further response is necessary.

   d) Variability versus Uncertainty

   Dr. Donna Vorhees

       The assessment incorporates discussion of correlations;  see response to charge
       question #9 for additional discussion regarding correlations.

       The assessment includes clear descriptions of distributions used in the PRA. However,
       the Authors made no attempt to differentiate  variability and uncertainty.  This level of
       effort is not necessarily warranted if conservative risk-based screening results are well
       below cumulative risk levels of concern.  Some distributions were truncated.  Truncation
       steps are not likely to strongly influence results of the analysis, and truncating at zero for
       inputs that cannot be negative is certainly reasonable as  long as one accounts for the


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       effect on parameters of the truncated distribution. Other truncation steps are not so
       easily defined. For example, is a reasonable maximum value for a homegrown produce
       ingestion rate really estimated by doubling the sum of the mean and 3*standard
       deviation? If not, what is the next best value? It seems far less complicated to leave
       distributions as they are with very low probabilities assigned to extreme values. The
       Authors could always use sensitivity analyses to examine the influence of extreme

       A puzzling aspect of the PRA  is the fact that only some inputs are defined with
       distributions. Why quantify variability and/or uncertainty for only some model inputs
       when data are available to develop distributions for others? The following sections
       describe information that is available for inputs that either were treated as point
       estimates in the PRA or were  defined with distributions that could be improved.

As implemented, the Monte Carlo simulation does not distinguish between uncertainty and
variability, and because the output distribution primarily represents the variability in the input
parameters, the Monte Carlo approach addresses what is generally referred to as Type A
uncertainty (the uncertainty associated with variability) (Hoffman and Hammonds, 1994). In
essence, distributions were selected to represent variability (e.g., exposure factors) and in some
cases, to also represent the uncertainty in the true parameter value (e.g., depth that manufactured
soil is incorporated into native soil). The uncertainty around the true mean (or other percentile
risk) and variance of the risk distribution are not addressed in the sense that the statistical
descriptors for each input parameter remain constant throughout the simulation, and the model
output is a single distribution of risk for each constituent. Type B uncertainty—the uncertainty
associated with a lack of knowledge—was not addressed separately from variability. Type B
uncertainty is distinguished from Type A uncertainty by conducting the Monte Carlo simulation
in two dimensions; this makes it possible to estimate a confidence interval around the probability
density function (PDF) and therefore allows the uncertainty in the overall result to be quantified.
The Authors have revised Sections 5.1 (Introduction to Probabilistic Modeling) and 6.8
(Uncertainty Characterization) to ensure that the distinction between uncertainty and variability
is made clear within the  context of this risk assessment.

The Authors agree that it is highly unlikely that the truncation of selected parameters will have a
significant influence on the risk results. The  Authors do recognize that the truncation of
parameter distributions is an ongoing  topic of discussion, and that truncation is only warranted to
prevent extreme (e.g., 1,500 Ib individual body weight) or impossible values (e.g., negative
numbers), as the commenter pointed out.  In response to the commenter's question regarding the
method chosen to establish the maximum value, this is based on a normally distributed
parameter. The value equal  to the mean plus 3 standard deviations is 99.865 (i.e., above the 99th
percentile). Recognizing that the distributional shape for most environmental exposure factors
approximates lognormal, the "protection factor" of 2 was added to ensure that a reasonable
maximum value is achieved. This approach to setting maximum values for Monte Carlo risk
assessment modeling was originally used during the development of the 3MRA modeling system
that was reviewed by EPA's Science Advisory Board
AB-05-003_unsigned.pdf),  and has been used by EPA in other multimedia risk assessments,


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
notably, the land application of biosolids. The Authors believe that this convention provides
reasonable high-end values for the distribution and, importantly, the maximum values were
assigned after the distribution was fit and the mean, standard deviation, and various percentiles
were determined. Although the "correct" value to assign as the maximum is a known unknown, a
quick comparison for one of the produce categories strongly suggests that this method does, in
fact, produce a reasonable maximum value that is highly unlikely to underestimate the "true"
maximum value. For instance, a 70 kg adult would consume 741 grams (WW) of exposed
vegetables per day using the maximum value calculated using this convention. This is roughly
the equivalent of eating two large salads every day that consist exclusively of vegetables grown
in a home garden, and is 341 grams higher than the recommended daily intake (400 grams per
day) of fresh vegetables, a mark that the vast majority of Americans fall well below. Given the
inherent conservatism in the approach described in the document (e.g., the average consumption
rate for home-grown produce was  105 grams [WW] per day), the maximum consumption rate
values are considered to be reasonable and appropriate for this risk assessment.

The Authors agree that, for certain types of risk assessments, the separation and quantification of
uncertainty and variability are desirable and necessary for the decision-making process.
However, for screening-level assessments that are designed to provide conservative estimates of
risk (i.e., the bias is designed to overestimate risk), the value of this additional information is
generally  not justified by the level of resources required to  develop the necessary input data, run
the model simulations, and analyze/present the results. For analyses (such as this SFS risk
assessment) that cover a significant proportion of the contiguous United States, it is difficult to
separate variability from uncertainty (Nauta, 2000), and a two-dimensional probabilistic
approach would have presented a real challenge  in terms of time and resources. Although a two-
dimensional Monte Carlo framework can provide additional insight into uncertainty by
separating variability and uncertainty, a one-dimensional probabilistic approach was used that
commingles variability and uncertainty into a single dimension (Mokhtari and Frey, 2005). Thus,
the risk distribution from the model simulation represents a  best estimate of the distribution of
risk for a unitized constituent concentration, accounting for  multiple sources of uncertainty and
variability, especially the variability in the input parameter values. For the purposes of screening
the potential for adverse health effects associated with the use of SFS-manufactured soils in
home gardens, the Authors do not believe that explicitly separating uncertainty and variability
would constitute a material improvement in the risk screening estimates.

Finally, the Authors point out that the PRA includes some inputs that are represented by single
values rather than point estimates.  Some are given explanations provided in the appendices (e.g.,
EPA-recommended values). In virtually all of the chemical  risk assessments conducted  over the
past 20 years, human health benchmarks (widely acknowledged as a significant source of
uncertainty) are represented by a point estimate rather than a distribution. Similarly, chemical
and physical properties are often given as a single best estimate, even though it is recognized that
there is variability associated with various measurement techniques. Moreover, some input
parameter values were chosen specifically to produce conservative estimates of potential health
risks,  which, as stated in the document, was the primary goal of the PRA. It should also be noted
that, because multimedia models such as 3MRA are sensitive to a relatively small number of
input  parameters (e.g., source concentration, ingestion rates), distributional data are not
developed for the entire suite of input parameters; this typically includes parameters to which the
model is not particularly sensitive, as well as parameters that exhibit relatively small variance.


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
Thus, given the objectives of the probabilistic risk screening, as well as value-of-information
considerations associated with separating uncertainty from variability, the Authors believe that
the models have been parameterized appropriately.

   e)  Inconsistencies with the Exposure Factors Handbook

   Dr. Mary Fox

       Some data inputs do not correspond to the source data referenced (see response to
       question 9, below)

       More care should be taken in defining minimum and maximum values on distributions
       used in the probabilistic modeling. For example, for the body-weight distributions, the
       mins and maxs found in Appendix I (Table 1-2) do not reflect the Exposure Factors
       Handbook data referenced (see comparisons below). It is especially important to choose
       conservative (and reasonable) maximums for probabilistic modeling particularly for body
       weight and averaging time, which appear in the denominator of exposure/dose
       equations. Generally speaking, when defining body weight and averaging time for a
       conservative scenario, lower values should be chosen. For greater transparency and
       reproducibility, inputs should reflect the source data.

       Table 1a. Comparing Data in Table 1-2 with EFH Data - Body Weight Minimums
Units = kg

Child 1
Child 2
Child 3
Min Table I-2

Min EFH Table
7-4 (5th%ile)

Min EFH Table
7-6 (5th%ile)
Min EFH Table
7-5 (5th%ile)

Min EFH Table
7-7 (5th%ile)
8.8- 15.3
32.2 - 48.5
       Table 1b. Comparing Data in Table 1-2 with EFH Data - Body Weight Maximums
Units = kg

Child 1
Child 2
Child 3
Max Table I-2

Max EFH Table
(95th %ile)

Max EFH Table
7-6 (95th %ile)
30.1 -61.0
Max EFH Table
7-5 (95th %ile)

Max EFH Table
7-7 (95th%ile)
       Note: EFH tables 7-2 and 7-3 are also referenced, but these contain data on means and
       not the tails of the distributions.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       Exposure duration - Table 1-2 lists the maximum value set for exposure duration at 100
       years - longer than the 70-year lifetime assumption reportedly used for cancer risk
       comparisons and longer than data in the table referenced (maximum value in EFH Table
       15-168 is 57 years).

       Table 1-3 Child 3 exposed fruit -1 believe there is a typo or calculation error.  On page I-
       3, it reads that the maximum was set at twice the 99th%ile. By my calculation that should
       be 5.9*2=11.8 g and not 18g.

   Dr. Donna Vorhees

       Cooking and Preparation Loss. Cooking and preparation loss data from USDA (Table 1
       in USDA [1975] Food yields summarized by different stages of preparation. Agriculture
       Handbook No. 102. U.S. Department of Agriculture, Agricultural Research Service,
       Washington, DC) could have been used to define distributions for this variable.  This
       publication is the source for the mean net cooking loss, mean net post-cooking loss, and
       mean paring and preparation loss (for fruits) values reported in Tables  13-6 and 13-7 in
       EPA's 1997 Exposure Factors Handbook.

       The exposure durations and averaging times are aligned. The PRA allows for variability
       in body weight,  and some toxicity values might incorporate a body-weight assumption of
       70 kg. If so,  I doubt that this inconsistency would have much influence on risk estimates.


The Authors describe how stochastic or distributed input data for each exposure factor were
collected and processed in Appendix I, Section 1.2.2. Exposure-related parameter values were
updated with data from the Child-Specific Exposure Factor's Handbook (CSEFH; U.S. EPA,
2008b) and the 2011 update to the Exposure Factors Handbook (2011EFH; U.S. EPA, 2011).
These data (i.e. from the CSEFH and 2011EFH) were used to fit distributions for Monte Carlo
analysis as described in this section. The minimum and maximum values are based on the
methodology developed for the 3MRA modeling system and, because the CSEFH and 2011EFH
data were used in fitting the distributions, the minimum and maximum values should not exactly
match those presented in the CSEFH and 2011EFH. Appendix I has been revised to ensure that
the basis for the minima and maxima is made clear. Similarly, the appendix discusses the basis
for the exposure duration distribution (see U.S. EPA, 2000) and corrects the typographical error
identified by the commenter for the exposed fruit maximum value. The Authors used the
recommended EPA values for cooking and preparation losses.

The Authors agree that the inconsistency in using health benchmarks derived assuming a 70 kg
adult with variable body weights does not have a significant effect on the risk estimates. This is a
widely acknowledged issue in the PRA.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   f)  Child-Specific Exposure Factors Handbook Not Used

   Dr. Mary Fox

       Section, page 5-15, Exposure Model Inputs
       The Authors should consult Child-Specific EFH to ensure they are using the currently
       accepted values for child intakes, etc. (In many cases, the data in the 1997 and more
       recent Child-Specific EFH may be the same.)

   Dr. Donna Vorhees

       More important, the risk assessment makes no reference to EPA's most recent guidance
       for evaluating childhood exposures ('Child-Specific Exposure Factor's Handbook,
       September 2008, Final). All analyses should be re-visited and updated as appropriate
       after considering the relevant data, analyses, and recommendations in this guidance

       Body Weight. The Authors use body-weight data from EPA's 1997 Exposure Factors
       Handbook. Body-weight data representative of the U.S. population have been collected
       more recently as part of the CDC's NHANES study. EPA developed distributions for
       children through age 21 using NHANES data from 1999-2006 (See Chapter 8 in EPA's
       2008 Child-Specific Exposure Factors Handbook;. Additional NHANES data could be
       obtained to evaluate adults.

       Soil Ingestion.  The Authors define soil ingestion with point estimates, but distributional
       information is provided in EPA's 1997 Exposure Factors Handbook, as well as the 2008
       Child-Specific Exposure Factors Handbook.


The Authors agree that new child-specific exposure data are now available. Also, in 2011, U.S.
EPA published an update to the Exposure Factors Handbook (U.S. EPA, 2011). As discussed in
6e above, exposure-related parameter values used in this assessment have been updated to reflect
data in the CSEFH and 2011EFH. Modeling was rerun and the report was modified accordingly.

   g) Data Collection Uncertainty

   Dr. Mary Fox

       Page 2-24, Discussion of TCLP and SPLP

       Usefulness of data "unresolved" potentially not representative of complex soil mixture
       settings. Is this an important uncertainty to include in uncertainty discussion? Is there
       any further information about this uncertainty (e.g., are the data expected to over- or
       underestimate contaminant concentrations from leaching in more complex settings) ?

   Dr. Charles Harvey

       / found the writing and organization in this section reasonably clear. The discussion of
       uncertainties should be broadened to include important uncertainties that are very
       difficult to assess from the available data. The section should discuss uncertainty
       associated with using 43 SFS samples to represent all SFS that would be provided by

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       large-scale projects.  The report section should also highlight the possibility of organic
       contaminants not considered in the assessment.

The Authors acknowledge that calling the usefulness of leachate data "unresolved" may confuse
readers. With respect to TCLP and SPLP, the Authors meant to say that simple leach tests cannot
capture actual leaching behavior under every conceivable set of conditions. The Authors do not
regard this as a significant source of uncertainty; however, because the aggressive leaching
conditions of TCLP and the acidic conditions represented by the SPLP provide conservative
estimates of the leaching  potential relative to the typical environmental conditions for the use of
SFS manufactured soils. For example, the home gardener would adjust the soil pH to near
neutral as  a normal part of growing home produce. The Authors have added the following
paragraph to Section 2.5.4 to address this concern:
       "The TCLP and SPLP represent standard tests that are widely used by the EPA
       and other regulatory agencies to evaluate the potential for constituent release into
       the subsurface. With  few exceptions,14 the aggressive conditions used in the
       TCLP described above are thought to provide a very conservative screen for leach
       potential. The scenario that the TCLP mimics, however, is not representative of
       SFS use in manufactured soil because the level of acidity will overestimate
       constituent release. In addition, the organic component of manufactured soils
       (e.g., composts, peat  moss, pine bark, biosolids) would likely sorb elements
       released from the  molding sand (Basta et al., 2005; Kumpiene et al., 2008). The
       SPLP conditions that mimic acid rain are more relevant than TCLP for evaluating
       the conditions considered in this report.
        14 Recent research indicates that the TCLP may not provide an adequately conservative test for
        arsenic in mature landfills characterized by alkaline pH, low redox potential, biological activity,
        long retention time, and organic composition of mature landfills (e.g., Ghosh et al., 2004)."

Concerning sample representativeness, a discussion is provided above under 3.d. Also, the
Authors have added the following sentence to Section 6.8.2 (Uncertainty Characterization, State-
of-the-Science on SFS), in a paragraph regarding the representativeness of available SFS data.

       "Nevertheless, it is unknown if the SFS samples from these 39 foundries are statistically
       representative of SFS from all iron, steel, and aluminum foundries. The related data may,
       therefore, overestimate or underestimate the range and  distribution of SFS constituent

Finally, the Authors have added the following text to Section 2.5.3 to address the potential for
additional organic contaminants.
       "While every effort was made to target the widest range of organic constituents that are
       of concern from an environmental and human health standpoint, it is possible that
       additional non-hazardous and hazardous organics were present in the SFSs and not
       addressed in this risk evaluation."

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   h) Toxicity Value Uncertainty

   Dr. Mary Fox

       Page 4-3
       Split Table 4-1 into sections for cancer and non-cancer benchmarks —add the health
       effect of concern for non-cancer benchmarks

   Dr. Donna Vorhees

       Except for lead and its associated CDC benchmark blood concentration, chemicals of
       concern selected in the assessment do not have toxicity values that are specific to
       susceptible subpopulations (e.g., PAH age-dependent adjustment factors). However, the
       Authors should discuss the potential for increased susceptibility among certain
       subpopulations in general, how they checked for this potential in evaluating risk from use
       of SFS in manufactured soil and other applications, and the results of their evaluation.

       Toxicity Values. EPA's PRA guidance:

          "does not propose probabilistic approaches for dose-response in human health
          assessment and, further, discourages undertaking such activities on a site-by-
          site basis" (EPA 2001).

       / assume that this is why EPA chose not to quantify uncertainty in toxicity values. But
       EPA should at least discuss uncertainty associated with chemical toxicity values in the
       risk assessment to facilitate interpretation of risk results.


With respect to Dr. Fox's request that Table 4-1 be broken down into cancer and noncancer with
health endpoints listed, the table has been modified.

Toxicity values were chosen based on the Office  of Solid Waste and Emergency Response
(OSWER) hierarchy (OSWER Directive 9285.7-53). The Authors acknowledge that toxicity
values are developed with uncertainty factors that account for variability among humans. While
this is not a perfect substitute, they are designed to account for sensitive subpopulations. A
discussion on uncertainty associated with chemical toxicity values has been added to Section
6.8.1 to facilitate interpretation of risk results. Further discussion of sensitive subpopulations is
given in 4.c, above.

   I)  Consumption Rate  Uncertainty

   Dr. Mary Fox

       / could not locate the data  on fraction  of home-grown produce grown in manufactured
       soil (home gardener scenario). This information is key to evaluating the conservatism of
       this scenario.

       Regarding the produce consumption modeling, the assessment uses consumption rate
       data from national surveys conducted in the late 1980s —this information is dated but
       remains in use.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       Gardeners will grow produce that they like and will consume it in season, as well as
       preserve it in various ways to be eaten in winter. Further, the probabilistic model inputs
       include no intake (minimums of 0 grams). The Authors do make a good point that a
       home gardener may not grow all five of the produce types, but likely grow 4 of the 5
       types. Another key consideration in evaluating the home gardener scenario is the
       fraction of produce assumed to be  home grown. I could not locate that number, so my
       evaluation of the conservatism of this scenario is incomplete.

   Dr. Donna Vorhees

       Assumption that the Consumption Rate for Homegrown Produce is 14 the General
       Population Consumption Rates.  This assumption is too simplistic. EPA (1997) provides
       some seasonally corrected consumption rates. Even where such adjusted data are not
       available, one can estimate adjustment factors that can be used to estimate seasonally
       adjusted consumption rates (See Section in Volume 5 ofEPA's 2005 Baseline
       Human Health Risk Assessment for the GE/Housatonic River Rest of River.,)

       Homegrown Produce Consumption Rates. A particular strength of this assessment is its
       reliance on recent research regarding plant uptake of metals for soil amended with SFS.
       Unfortunately, this research is ultimately combined with consumption rate data for home-
       produced food that is nearly 20 years old. EPA (1997) cautions those who use these
       data that they may be outdated, but the Authors of this assessment are silent on this
       topic. Unfortunately, I am not aware of more recent, systematically collected
       consumption rate data for home-produced food that are representative of the U. S.
       population. However, I have attached a recent National Gardening Association (NGA,
       2009) white paper that suggests that home gardening is on the rise. I am not familiar
       with the NGA survey beyond this report and cannot attest to its accuracy. Plus, it looks
       forward rather than backward in time and does not provide consumption rate data
       needed to quantify exposure. However, findings from the report mirror a trend that I've
       observed anecdotally in the northeast and suggest that the uncertainty associated with
       20-year old consumption rate data  warrants at least some discussion in the assessment.
       The Authors refer to the consumption rates as "conservative" based on comparison to a
       1993 USDA risk assessment, but this comparison is irrelevant. In addition, the Authors
       argue that

          "In the probabilistic modeling conducted for this assessment, the total
          consumption rate of home-grown fruits and vegetables for the adult at the 90th
          percentile risk level was approximately 500 g (WW) d"1 for an average adult.
          In addition, it is not possible to  harvest most garden crops for more than  a
          short period when the crop is ripe, which considerably limits potential
          exposure to garden foods. Given the size of the garden required to support
          such a diet, the  costs of delivering SFS would likely reduce the actual
          exposure to manufactured soil containing SFS by  several orders of magnitude
          due to the limited garden area. Thus, the results of the home gardener risk
          screening modeling should be considered as a significant overestimation of
          the actual risks  associated with  SFS use."

       The NGA white paper reports that the average size of a home garden is 600 square feet
       and that a well-tended garden produces 14 pound of produce per square foot, or about
       300 pounds per year. This equates to about 380 g/d fora 1-person household or about


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       90 g/d for a 4-person household. Again, these quantities are based on mean garden
       size, not upper-percentile garden sizes, and they do not include consumption of produce
       grown on other agricultural land where SFS might be used. I imagine authors from the
       USDA would have additional and perhaps better sources of data to estimate this
       quantity, and I strongly recommend that this discussion be included. Consumption
       should also be described in terms that people understand. For example, the 90th
       percentile consumption rate of 500 g/day corresponds  to 2 or 3 garden tomatoes per
       day. (Note: the garden size assumed in the assessment is 111-180 acres, while the
       NGA [2009] reports that 6% of home gardens are greater than 2,000 square feet without
       specifying a maximum value. Nevertheless, as the Authors note, their assumption of
       garden size greatly overestimates the size of home gardens.)

       Future assessments of SFS or other materials proposed for beneficial use should extend
       beyond research regarding environmental mobility and uptake and include studies to
       improve our understanding of important human exposure variables, such as
       consumption rates of homegrown produce.


With respect to the fraction of homegrown produce grown in manufactured soil, the Authors
have modified Section 3.1.4 to be clearer about the scenario.
       "Although manufactured soil could be used in corporate and residential
       landscaping (e.g., resurfacing construction sites), the home gardener would
       potentially receive a much higher exposure to SFS constituents under the
       following assumptions
       •   The home gardener incorporates a significant amount of manufactured
          soil into the home  garden
       •   The home gardener frequently works in the garden, thereby increasing
          the opportunities of dermal contact and incidental ingestion of SFS
          manufactured soil, and
       •   A significant portion of produce consumed by the home gardener
          would be taken from the garden consisting of SFS manufactured  soil."

With regards to the fraction of produce assumed to be homegrown,  the Authors point to Sections and, pages 5-14 through 5-15, of the original risk assessment. There, it is stated
       "Exposure through the ingestion route was estimated by multiplying the
       concentration of the constituent in the soil or food item by the consumption rate of
       the individual. [...] USDA was concerned that the distributions used to estimate
       consumption rates might result in overly conservative consumption rates of home-
       grown produce. To further investigate this, two additional sets of runs were added
       for comparison: one using point estimates of 50th percentile  annual produce
       consumption rates for  the general population, multiplied by  50% to account for
       crop growth periods and climate limitations to crop harvest periods (reducing the
       effective consumption rate to home grown produce); and a set of runs using the
       90th percentile annual produce consumption rates for the general population,
       similarly multiplied by 50%."


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
U.S. EPA (2011) provides "homegrown consumption" rates that represent the quantity of
produce consumed from the home garden. This may or may not represent 100% of total
consumption of these items, but clearly represents the amount of home-grown produce
consumed. Thus, the issue of "proportion" regarding the amount of produce consumed from the
home garden is not relevant; these consumption rates represent the amount of produce grown in
the home garden that is consumed; that is, the consumption rates are specific to the fraction of
the diet that is from the home garden and do NOT reflect total consumption of produce. The
consumer could be eating additional produce from other sources, but, because the scenario is
defined for home grown gardens and not commercial gardens, additional exposure to
constituents in SFS is presumed not to occur (i.e., the person is assumed NOT to collect and eat a
significant amount of produce from other home gardens). The Authors have revised the bulleted
text description of the three sets of modeling runs in Section to clarify this point, as

       •  "Set 1: Home gardener, modeled distributions of consumption rates (for home
          gardeners) - the produce consumption rates specific to home grown produce;
       •  Set 2: General population, 50th percentile (for the general population)
          consumption rates - the median produce consumption rates for the general
          population were multiplied by 0.5 to derive a value specific to home grown
          produce; and

       •  Set 3: General population, 90th percentile (for the general population)
          consumption rates - the high end produce consumption rates for the general
          population were multiplied by 0.5 to derive a value specific to home grown

Dr. Vorhees suggests that an alternative to the 50% home-grown rate used for the general
population scenarios would be to estimate seasonally adjusted consumption factors as was done
in U.S. ACE and U.S. EPA (2005). However, for a screening-level risk assessment intended to
support decisions involving SFS across a significant portion of the contiguous United States (see
Figure 3-5), the use of seasonally adjusted consumption factors would introduce additional
uncertainty and provide little value given the level of resources required to  develop seasonal
consumption rates, modify the model code to derive seasonal estimates for  biotransfer factors,
and re-run the simulations. The Authors believe that the use of empirical soil-to-plant biotransfer
factors for broad categories of produce  (e.g., exposed vegetables) represent a much greater
source of uncertainty than the simple 50% adjustment provided by USDA to support  a
comparative set of modeling runs. More importantly, EPA continues to recommend the use of
EFH2011 data in the absence of newer  study data. In fact, an earlier trend analysis that used data
from the state-based Behavioral Risk Factor Surveillance System (BRFSS) found that fruit and
vegetable consumption by American adults was essentially unchanged from 1994 through 2000,
and that a low proportion of Americans ate five3 or more fruits and vegetables per day (Serdula
et al., 2004). As discussed in the problem formulation, the purpose of the screening risk
assessment was to evaluate the potential for adverse health and ecological effects  associated with
the use of SFS in specific manufactured soil applications (e.g., incorporation into home gardens).
The Authors acknowledge that there are sources of uncertainty in the screening risk assessment
3 This equates to approximately 400 grams per day for an average adult.


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
(as with any risk assessment) and that additional data development activities and modeling could
be performed to reduce those uncertainties. However, the Authors would like to emphasize that
the assumptions and parameterization of the model were intentionally conservative to ensure that
potential risks would not be underestimated. Thus, additional activities would only be undertaken
if they materially improved the quality of the information used to support the decision-making
process for the states. The Authors firmly believe that the screening level risk assessment is
appropriately conservative (i.e., fit for purpose) to support defensible conclusions regarding the
use of SFS in manufactured soils.

Regarding Dr. Vorhees' suggestion of the National Gardening Association 2009 white paper as a
useful reference, the Authors appreciate the suggestion and have used the suggested paper to
inform modifications to the home  garden conceptual model (see report Section 5.2.1).

Finally,  as discussed in 5.d above, the Authors have replaced the phrase "significant
overestimation" with the phrase "as an overestimate" in the revised document.

   j)  Cumulative Risk

   Dr. Mary Fox

      Section 4.1, page 4-1

      More justification is needed to support separate (not cumulative) evaluation of pathways.
      Inhalation and ingestion -  what are the critical health effects underlying health
      benchmarks for each constituent of concern for each route of exposure? Ingestion: What
      is known about leaching to groundwater? How long does it take? Quantify/describe the
      difference in time-scale. Inhalation and ingestion in combination would also seem
      plausible for residents near a soil blending plant.

      Section 4-4, page 4-11

      Some SFS constituents do not have tox benchmarks (so they are not included in the
      assessment) - is this discussed as possible source of underestimation of risk in
      limitations or uncertainty section?

   Dr. Charles Harvey

      Page ES-3, paragraph 2

       The assessment should document the claim that inhalation and ingestion cause different
      health impacts -1 was not aware that this  is true across the range of contaminants
      considered here. Furthermore, the effects of ingestion on different time scales could be
      cumulative. For example, I am unaware of any research that indicates arsenic ingestion
      over different timescales is not cumulative. I suspect that rapid exposure from produce
      followed later by exposure from groundwater could be cumulative.

       The assessment should document the claim of independence of ingestion pathways. I
      am not aware (across the range of contaminants) that inhalation and ingestion cause
      different health impacts (e.g., lead?). Furthermore, the effects of ingestion on different
      time scales could be cumulative, so groundwater and produce may not be independent
      pathways. For example, I am unaware of any research that indicates arsenic ingestion
      over different timescales is not cumulative.


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   Dr. Donna Vorhees

       / assume that "appropriately conservative" means that cumulative noncancer hazard
       indices do not exceed 1 and cumulative cancer risks do not exceed 1E-5 for any
       receptors. In all likelihood, the screening steps were appropriately conservative, but, as
       explained in response to charge questions #1 and #6 [see comments below], the
       Authors should clearly and succinctly document quantitatively how screening steps
       throughout the report ensure that SFS use will not be associated with cumulative risk
       levels of concern.

       Specifically, demonstrate briefly but quantitatively why SFS use is not associated with
       cumulative risk levels of concern despite:

       1.  Screening out chemicals that were never detected (this step should not be
          problematic because the authors checked for and addressed detection limits that
          exceeded screening levels),
       2.  Screening out chemicals that do not have health benchmarks,
       3.  Assuming independence among some exposure pathways (exception is the
          evaluation of cumulative ingestion exposure to soil and homegrown food),
       4.  Applying the target hazard index of 1 to single chemicals associated with each
          exposure pathway despite the fact that each exposure pathway involves exposure to
          chemical mixtures,
       5.  eliminating some exposure pathways from  quantitative evaluation (e.g., dermal
          contact with soil and groundwater and inhalation of fugitive dust [although predicted
          soil screening concentrations for the dust pathway are sufficiently high relative to
          SFS concentrations that this pathway should not contribute negligibly to cumulative
       6.  Use of 95th percentile concentrations instead of maximum detected concentrations to
          screen for chemicals of concern (it is common practice in EPA's Superfund program
          to use the maximum detected concentration for this purpose, but practices might
          vary among different federal and state programs), and
       7.  assuming that the groundwater ingestion pathway did not require further evaluation
          because estimated exposures for five modeled constituents were below EPA's MCLs
          [While this assumption may be practical, MCLs are not all necessarily risk-based].

       This comment is related to the  graphic suggested in response to charge question #1.
       More attention needs to be paid to the concept of cumulative risk, referencing relevant
       EPA guidance (e.g., U.S. EPA  2000, 2003,  2007).

       The Authors assumed independence of groundwater, soil, and fugitive dust exposure
       pathways for the following reasons: "Each of the three pathways listed above was
       evaluated through a screening  model to see if any pathway (or alternatively, any
       constituents) could be eliminated from further analysis. It is important to note that these
       pathways are likely to operate individually on a human receptor, not cumulatively. First,
       inhalation of materials will generally cause different health impacts than ingestion of
       those materials. Therefore, the inhalation pathway should be evaluated separately from
       the ingestion pathways. Second, exposures via groundwater ingestion occur on a
       significantly different time-scale from ingestion of produce and soil. Thus, the
       groundwater pathway can also be evaluated separately. Given the individual nature of
       these pathways, they were each evaluated in turn." The Authors provide no justification
       for assuming that the fugitive dust pathway and ingestion pathway for soil should be


Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       separate. Would one really expect different health effects for the COCs in this
       assessment? What about the fraction of fugitive dust that is ultimately ingested rather
       than inhaled (i.e., that fraction that enters the airway and is cleared via the mucociliary
       escalator before entering the gastrointestinal tract) ? I will leave it to those who are expert
       in groundwater modeling to comment on the timescales, but it seems that exposures to
       SFS in groundwater and soil could occur at the same time and place if SFS-containing
       materials are used in the same place over time.

The Authors acknowledge that some chemicals may interact in a mixture, causing different
health outcomes than would result from individual exposures. Additionally, the Authors
acknowledge that individuals may be exposed to the same chemical through multiple pathways.
Since additive risk across constituents and pathways is a possibility,  the Authors have addressed
each of the commenters' concerns below, beginning with Dr. Vorhees' seven enumerated points.

Regarding Dr. Vorhees' first point, that constituents were screened out when not detected, the
Authors point out that these constituents were addressed when the detection limits exceeded
screening levels.

Dr. Vorhees' next point concerns chemicals that do not have health benchmarks, a comment
echoed by Dr. Fox. The Authors note that EPA continually  strives to assess health benchmarks
for a growing array of constituents and mixtures. However,  without further toxicological work,
which is beyond the scope of this assessment, the Authors cannot evaluate risks from these
constituents and mixtures. The Authors have modified the language in several places to
transparently state the limitations of the assessment, including the quantitative evaluation only of
those constituents for which benchmarks are available, and  to explicitly state the uncertainty
inherent in this limitation. The Authors have also added the following text to Section
       "The exposure scenarios and pathway evaluations were developed to produce
       highly conservative estimates of risk; that is, the methodology was designed to
       overestimate the actual risk to ensure that an ample margin of safety was built into
       the analysis."

The third potential concern Dr. Vorhees raises, as does Dr. Harvey, is the Authors' assumption
of independence across exposure pathways.  EPA agrees that the rationale in the report does not
adequately explain the relevance of focusing on three basic  pathways, and why cumulative risks
were not fully evaluated. The Authors have  modified which Soil Screening Levels (SSLs) are
used in Phase I: The assessment now uses Residential SSLs that, on  a constituent-specific basis,
can address two or more exposure pathways (i.e., in addition to soil ingestion, they can also
address dermal exposure to soil, inhalation of fugitive dust,  or both). Also, additional analysis
was performed, and  documented in report Section and Appendix J, demonstrating that
surface pathway exposures and groundwater exposures would not occur in the same time-frame.
Refined modeling therefore did not aggregate surface pathway and groundwater exposures. The
Authors have revised several sections in the report to clarify how the assessment addressed
cumulative risks in Phase I, while focusing Phase II modeling on three basic exposure pathways
identified during the development of the exposure scenarios (as shown in the conceptual
models). To summarize

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       Phase I screening used the SFS data and available screening criteria and models to
       determine which constituents, if any, should be considered for probabilistic risk
       modeling. Inhalation exposure and groundwater exposure were screened individually.
       However, soil pathways screening also included dermal and inhalation exposures, to the
       extent that constituent-specific data were available. If the constituent failed any one of the
       screening steps, it was subjected to a more rigorous modeling approach to further screen
       the constituent on the basis of potential health or ecological risk. Even aggregating
       exposures, Phase I screening demonstrated that no constituents required refined
       inhalation or dermal modeling. Phase II modeling focused, therefore, on human exposure
       via incidental soil ingestion and ingestion of home-grown produce, and ecological
       exposures via direct contact. Additional analysis demonstrated that groundwater
       exposures and surface pathway exposures would not happen in the same time-frame, and
       these exposures were therefore evaluated separately.
       Although some constituents can elicit similar toxicological responses (e.g.,
       neurotoxicity), neither the screening nor the modeling stages of the analysis aggregated
       exposures across multiple constituents. The Authors consider the overall design of the
       assessment sufficiently conservative as to make further assessment of cumulative risk
       unnecessary. For example, the exposure scenarios and pathway evaluations were
       developed to produce highly conservative, reasonable maximum exposure estimates of
       risk, to ensure that an ample margin of safety was built into the analysis. This approach
       ensures that the results of this analysis can be used to confidently determine if soil-related
       uses of SFS will be protective of human health and the environment. The risk assessment
       is therefore an efficient approach to providing decision makers with information on the
       potential for adverse effects to the most highly exposed individuals and ecological
       receptors that could come in contact with SFS constituents.

Dr. Vorhees is correct in her next point, assessing that exposures modeled in this assessment are
chemical mixtures, and not individual chemicals. Although HQs of 1 are only for individual
chemicals, the quantitative human health benchmarks available to the Authors are also based on
the toxicity of individual chemicals. Thus, without the further research into risks of combinations
of chemicals, this approach will continue to pose the potential to underestimate or overestimate
risks. This uncertainty is now discussed in more detail in Section 5.3.8 Human Health Effects
Modeling. Additionally, the Authors note that it is possible for one foundry sand to have higher
concentrations of constituent A and lower concentrations of constituent B when compared to
another foundry sand. This creates difficulty in conducting a unitized model effort and
developing screening levels for individual constituents as was done here.

The fifth point in Dr. Vorhees' list of cumulative risk considerations is that the Authors do not
include dermal and inhalation exposures in cumulative risk estimates. However, as the
commenter correctly points out, the inhalation screening values are orders of magnitude higher
than those for ingestion, and thus are unlikely to contribute significantly to overall risk. Also, as
discussed in Section 4f, the Authors have performed additional screening evaluations of dermal
exposures which, like the inhalation evaluation, included screening values at least an order  of
magnitude higher than those for ingestion.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

The next suggestion Dr. Vorhees proposes is that the Authors use maximums instead of 95th
percentile concentrations. The Authors first point out that maximums were used for dioxins,
furans, and PCBs, and that maximums for all elements are presented in the final summary tables
alongside the 95th percentile. For the remaining constituents, many were not detected in any
sample, and thus would not have changed the evaluation. For those constituents that were
detected but passed the screen at the 95th percentile, only one would not have passed the screen
at the maximum. Zinc would have failed for ecological risk (123 mg kg"1 versus an EcoSSL of
120 mg kg"1). However, as discussed in report Appendix C, a zinc screening level of 300 mg kg"1
is protective of soil fertility. Thus, it would be unnecessary to model zinc.

Dr. Vorhees seventh and final enumerated consideration for cumulative risk is that the potential
for cumulative risk in the groundwater pathway should not be screened away by comparing to
the MCLs because they are not health-based levels. The Authors agree that MCLs are not
necessarily based solely on risk. For precisely this reason, the Authors also compared leachate to
tapwater screening levels from the regional screening tables.

The Authors agree with Doctors Fox, Harvey, and Vorhees  that greater justification was needed
for the assumption that exposures via surface pathways (i.e., incidental ingestion of garden soil,
and ingestion of home-grown produce) would not occur in the same timeframe as exposures via
groundwater. As discussed in 5c above (and documented in report Section 5.3.5 and Appendix
J), additional modeling was performed that validates and quantifies this assumption.

Finally, the Authors wish to reiterate that concentrations of most constituents in SFS are below
the same concentrations in naturally occurring soils. Thus, the cumulative risks experienced here
would differ very little from those posed by native soils.

   k)  Clarity

   Dr. Mary Fox
       / think Chapter 6 contains most of the relevant information to characterize the
       assessment and put the results in context. However, the clarity is compromised by some
       organizational issues—there are some sections that appear out of place and some
       sections that don't seem well integrated into the discussion at all.

       Section 6.2, Key risk assessment questions
       The fourth question (nutritional health and essentiality)  doesn't seem to be directly
       addressed in the chapter.

       Section 6.3.5
       This section is not well-integrated into Chapter 6. How does this discussion of highly
       exposed populations relate to uncertainty or the assessment overall? Does it relate to
       how ecological risks were evaluated?

       Section 6.4
       The information in Section 6.4 seems more appropriate as part of the preceding section
       on Overarching Concepts.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       Section 6.5 through 6.7
       These substance-specific sections are good summaries of the assessment information. I
       would substitute "Summary" or "Integrated Summary" for "Weight of evidence" because
       weight of evidence is risk assessment jargon that can mean different things to different

       The Authors present Section 6.8, Uncertainty Characterization, as a high-level overview
       for risk managers/policy makers and therefore do not re-hash assumptions and
       uncertainties in detail. The information presented is useful; however, as a technical
       reader, I was looking for more. I would like to see a "Data and Research" section where
       the Authors comment on data quality, data gaps, and the feasibility/desirability of a
       validation study or other research needs.

The Authors acknowledge that the discussion in Chapter 6 could be clearer. The Authors believe
that changes made throughout the report and, specifically, in Chapter 6, improve the clarity and
accessibility of information in this chapter. However, the Authors do not believe that additional
changes are warranted, as suggested by the following responses.
   •   While nutritional health and essentiality are not discussed as prominently as the risk
       results, they do appear throughout the chapter. For example, in Section the
       nutritional role of manganese is discussed.
   •   As discussed in 6.a, above, "weight of evidence" has been changed to "lines of
   •   The Authors believe that the data quality and data gaps have been examined in detail in
       the risk characterization and that the risk assessment that was developed to support safe
       management  of SFS in manufactured soils is appropriate and adequate for the purpose.
       Ongoing work may identify additional research needs as they pertain to other applications
       of SFS.

7) General/Other

The peer-review comments not related to the previous categories are grouped into four
   a)  Non-Technical Abstract and Public Label;
   b)  Technical Inaccuracies/Editorial Comments;
   c)  Application to States; and
   d)  Risk Assessment versus Risk Management.

   a)  Non-Technical Abstract and Public Label

   Dr. Ken Barbarick
       The American Foundry Society's request for an Abstract (I would recommend 1 page or
       less) and their suggested final statement at the bottom of page 2 of their response are
       reasonable requests. I also support their recommendation to call the material "recycled

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       foundry sand." This change puts a more positive spin on the nature of the material and
       how it could be re-used.
The Authors agree that an abstract or non-technical summary of this document would be useful.
While outside the scope of this report, the Authors will consider providing such a document in
the near future.

   b)  Technical Inaccuracies/Editorial Comments
   Dr. Ken Barbarick
       The American Foundry Society's comments point out some technical inaccuracies
       concerning the foundry processes and materials that should be corrected.
   Dr. Mary Fox
       Page ES-7
       Statement is made that composition of SFS may reduce bioavailability of lead, but no
       reference is provided.
       Section 3.1.5, page 3-6
       Assumptions on indirect exposure pathways from temporary storage and use of SFS—
       reference needed to support the claim about biomagnification
       Page 4-16, Table 4-9
       Adjustment to the SSL should be presented. Why is SSL for lead shaded in gray?
       Page 4-17, first paragraph, sentence 4
       Check spelling for 'arsenic'
   Dr. Charles Harvey
       Page 1-4, paragraph 5
       Needs editing -"... the characteristics of individual metals, such as the soil-plant
       Page 3-13, paragraph 1
       Needs editing - "It was also clear that certain scenarios were more significant in some
       scenarios than in others."
   Dr. Donna Vorhees
       The document is generally clear and well-organized, although it would benefit from more
       succinct text in some places. For example, the statement that SFS is assumed to
       comprise 50%  of manufactured soil appears 11 times in Sections 1 through 5 alone.

Responses to Review Comments     Risk Assessment of Spent Foundry Sands in Soil-Related Applications

The Authors acknowledge the technical inaccuracies pointed out by the American Foundry
Society and have corrected these in the revised risk assessment.

The statement regarding the biomagnification of chemical constituents in terrestrial food webs
refers to the lack of published studies that demonstrate this phenomenon. With the exception of
certain persistent organic pollutants, such as dioxins and PCBs, the Authors are not aware of any
studies demonstrating biomagnification for multiple trophic levels (e.g., from soil invertebrates
up through top predators). The Authors have clarified the related language.

The SSL for lead was not adjusted. Thus, it should not have been gray. The Authors have
corrected this error in the revised risk assessment.

The misspelling of arsenic on former page 4-17 has been corrected in the revised risk

The sentence on former page 1-4 has been edited to be clearer. As discussed previously, the
Analysis Plan section 3.2 has been rewritten and the sentence noted on the former page 3-13 is
no longer in the document.

Finally, the Authors have attempted to reduce statements regarding the assumption that SFS
make up 50% of the manufactured soils modeled.

   c) Application to States

   Dr. Ken Barbarick

       The report submitted by the Michigan Department of Environmental Quality studied the
       data for possible impacts and concluded that the material possibly could be used with
       restrictions. I believe the report actually adequately addressed the issues raised by the
       Michigan Department of Environmental Quality.

   Dr. Donna Vorhees

       The PRA modeling steps are generally appropriate to develop national-scale screening
       values. However, more work is needed to comply with EPA  recommendations for PRA
       documentation and EPA's most recent recommendations for evaluating children's
       exposure. In addition, the Authors should consider adding a section that explains how
       states might modify the analyses to incorporate state-specific information, thus reducing
       the uncertainty in applying results of a "national-scale" model to specific locations.

       Application of National-Scale Screening Values and PRA to Individual Regions/States.
       The assessment accounts for variability in chemical mobility in the environment and in
       soil background concentrations across the United States. This accounting of regional
       variability is essential fora national-scale analysis. To provide states with as much
       flexibility as possible in applying findings in a manner that ensures compliance with their
       own risk management goals, the Authors could include a section explaining how states
       could substitute their own data (e.g., soil characteristics and chemical concentrations)
       into the PRA model and other screening models.


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
       In addition, some states have defined soil background concentration data sets that could
       be presented in this assessment along with the SFS data, USGS data, and other data
       briefly discussed in Section 6.8.2, item #2 (e.g., see "Background Levels ofPolycyclic
       Aromatic Hydrocarbons and Metals in Soil" at
       http://www. mass. Qov/dep/service/compliance/riskasmt. htm).


The Authors acknowledge that a site-specific or state-specific risk assessment could be
conducted for the beneficial use of SFS in the future. Including guidance on how states could use
the methodology to conduct state-specific evaluations is beyond the scope of this report.
However, by being as transparent as possible in describing the methodologies and data used in
this assessment, it was the Authors' intention to facilitate future application of the methodology
by other interested parties. For example, the Authors point to the fact that states can easily use
totals sampling and leachate testing to compare their actual SFS to the 95th percentiles that were
found not to pose excess risks in this report. In addition, as Dr. Vorhees correctly points out,
some states have already developed state background soil levels for comparison. While the
Authors encourage states to make use of the best, most specific data available to them, and will
continue to  assist states in future analyses, such work is outside of the scope of this document.
However, as described in 7.d, below, this may be addressed through a non-technical fact sheet
and training.

   d)  Risk Assessment versus Risk Management

   Dr. Ken Barbarick

       The EPA Region 9 comments point out the mixing of Risk Assessment and Risk
       Management approaches.  I agree that this needs clarification and the report should
       focus on Risk Assessment.

   Dr. Charles Harvey

       The study arrives at a strong conclusion (ES-8): "...no evidence was found that the
       specified uses of non-olivine SFS produced by iron, steel, and aluminum foundries
       evaluated in this report could pose significant risks to human health or the environment
       when used in manufactured soils, soil-less media, or road base." This statement (and
       other statements) is more than objective descriptions of the risks of using SFS; it is a
       value judgment about whether the risks of the anticipated uses of SFS are acceptable.
       As such, the conclusion combines both a quantification of the risks and an assessment
       of whether these risks are acceptable.  The document would be easier to follow if clearly
       separated these two steps. However, I was not convinced that the study fully considered
       the second step, the judgment that risks are acceptable. For example, would the risks be
       acceptable under all types of SFS use? Is the choice of the 90-percentile risk
       appropriate, or should risks in the top decile, that are potentially much higher, also be
       considered? Do the risks need to be weighed against the benefits?

       I think the document would be easier to follow, and would remain just as valuable, if it
       simply stated the purpose of providing a good assessment of the risks and then adhered
       to this narrower purpose.

Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
   Dr. Donna Vorhees
       There is a tone of advocacy at several points in this document that are not typically
       found in risk assessments, nor are they helpful as they stray from the topic at hand. For
       example: "Given their inherent properties and low cost, SFSs present a significant
       opportunity for the manufacture of soil and soil-less media" (Page 3-4).  From a technical
       perspective, it appears that the work was performed in a scientifically objective manner,
       but such statements do not instill confidence that the risk evaluation was conducted
       objectively in the minds of those who are unfamiliar with the details of the technical
       evaluation. I suggest that the Authors consider deleting them.

The Authors have modified several sentences in the report that could be interpreted as  evidence
that the analysis was not performed in a scientifically objective manner. However, it should be
pointed out that the analysis was intended to provide an objective description of all properties
and characteristics of SFS, including those that can be valuable in soil manufacturing and

8) References
Blanck, H.M.; Gillespie, C.; Kimmons, I.E.; Seymour, J.D.; Serdula, M.K.2008. Trends in fruit
       and vegetable consumption among U.S. men and women, 1994-2005. Preventing
       Chronic Disease 5(2). Available at http://www.cdc.gov/pcd/issues/2008/apr/07_0049.htm
       (accessed 30 June 2014).
Ghosh, A., M. Mukiibi and W. Ela. 2004. TCLP underestimates leaching of arsenic from solid
       residuals under landfill conditions. Environ. Sci. Technol. 38:4677-4682.
lEc (Industrial Economics). 2009. Peer review of Risk Evaluation of Spent Foundry Sands in
       Soil-Related Applications. Memorandum prepared for Richard Benware; U.S. EPA,
       Office of Resource Conservation and Recovery. Washington, DC. November.
Ritchie, N.J., M.E. Schutter, R.P. Dick and and D.D. Myrold. 2000. Use of length heterogeneity
       PCR and fatty acid methyl ester profiles to characterize microbial communities in soil.
       Appl. Environ. Microbiol. 66:1668-1675.
Schutter, Sandeno, and Dick. 2001. Seasonal, soil type, and alternative management influences
       on microbial communities of vegetable cropping systems. Biology and Fertility of Soils
Serdula, M.K., C. Gillespie, L. Kettel-Khan, R. Farris, J. Seymour and C.  Denny. 2004. Trends
       in fruit and vegetable consumption among adults in the United States: Behavioral risk
       factor surveillance system, 1994-2000. Am. J. Pub. Health 94:1014-1018.
U.S. ACE (Army Corps of Engineers) and U.S. EPA (Environmental  Protection Agency). 2005.
       Baseline Human Health Risk Assessment for the GE/Housatonic Rest of River Site. GE-
       021105-ACMT. February.
U.S. Census Bureau. 2002. 2002 Economic Census. Table 1 NAICS code 3315.
U.S. EPA (Environmental Protection Agency). 1989. Risk Assessment Guidance for Superfund
       (RAGS) Part A. EPA/540/1-89/002. Office of Emergency and Remedial Response,


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications

       Washington, DC 20450. December. Available at
       http://www.epa.gov/oswer/riskassessment/ragsa/index.htm (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 1991. EPA Region 3 Guidance on Handling
       Chemical Concentration Data Near the Detection Limit in Risk Assessments. Region 3,
       Philadelphia, PA. November. Available at
       http://www.epa.gov/reg3hscd/risk/human/info/guide3 .htm (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 1995. SCREENS Model User's Guide. EPA-
       454/B-95-004. U.S. Environmental Protection Agency, Office of Air Quality Planning
       and  Standards, Emissions, Monitoring, and Analysis Division, Research Triangle Park,
       North Carolina Available at http://www.epa.gov/scramOO 1 /userg/screen/screen3d.pdf
       (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 1997a. Exposure Factors Handbook, Volume I,
       General Factors. EPA/600/P-95/002Fa. U.S. Environmental Protection Agency, Office
       of Research and Development, Washington, DC. August. Available at
       http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid= 12464 (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 1997b. Exposure Factors Handbook, Volume II,
       FoodIngestion Factors. EPA/600/P-95/002Fb. U.S. Environmental Protection Agency,
       Office of Research and Development, Washington, DC. August.  Available at
       http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid= 12464 (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 1997c. Exposure Factors Handbook, Volume III,
       Activity Factors. EPA/600/P-95/002Fa. U.S. Environmental Protection Agency, Office of
       Research and Development, Washington, DC. August. Available at
       http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid= 12464 (accessed 30 June 2014)
U.S. EPA (Environmental Protection Agency). 1999a. Contract Laboratory Program Statement
       of Work for Inorganic Analysis, Multi-Media, Multi-Concentration. Document number
       ILM04.0b. U.S. Environmental Protection Agency, Washington, DC.
U.S. EPA (Environmental Protection Agency). 1999b. Air Module Pre- and Postprocessor:
       Background and Implementation for the Multimedia, Multipathway, andMultireceptor
       Risk Assessment (3MRA) for HWIR99. U.S. Environmental Protection Agency,  Office of
       Hazardous Waste, Washington, DC. October. Available at
       (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 2000. Supplementary Guidance for Conducting
       Health Risk Assessment of Chemical Mixtures. EPA/630/R-00/002. U.S. Environmental
       Protection Agency, Risk Assessment Forum, Washington, DC. August. Available at
       http://www.epa.gov/raf/publications/sup-guidance-hra-chem-mix.htm (accessed 30 June
U.S. EPA (Environmental Protection Agency). 2002a. Industrial Waste Management Evaluation
       Model (IWEM) Technical Background Document. EPA530-R-02-012 U.S.
       Environmental Protection Agency, Office of Solid Waste and Emergency Response,
       Washington, DC. August. Available at
       http://www.epa.gov/epawaste/nonhaz/industrial/tools/iwem/index.htm (accessed 30 June


Responses to Review Comments    Risk Assessment of Spent Foundry Sands in Soil-Related Applications
U.S. EPA (Environmental Protection Agency). 2003 a. EPA 's Composite Model for Leachate
      Migration with Transformation Products (EPACMTP): Parameters/Data Background
      Document. Office of Solid Waste, Washington, DC. April. Available at
      http://www.epa.gov/osw/nonhaz/industrial/tools/cmtp/index.htm (accessed 30 June
U.S. EPA (Environmental Protection Agency). 2003b. EPACMTP Technical Background
      Document. Office of Solid Waste, Washington, DC. Available at
      http://www.epa.gov/osw/nonhaz/industrial/tools/cmtp/index.htm (accessed 30 June
U.S. EPA (Environmental Protection Agency). 2003c. Multimedia, Multipathway, and
      Multireceptor Risk Assessment (3MRA) Modeling System Volume I: Modeling System and
      Science. U.S. Environmental Protection Agency, Office of Solid Waste, Washington DC.
      530-D-03-001a. Available at
      http://www.epa.gov/osw/hazard/wastetypes/wasteid/hwirwste/risk03.htm (accessed 30
      June 2014).
U.S. EPA (Environmental Protection Agency). 2004. EPA's Multimedia Multipathway and
      Multireceptor Risk Assessment (3MRA) Modeling System; A Review  by the 3MRA Review
      Panel of the EPA Science Advisory Board. U.S. Environmental Protection Agency,
      Science Advisory Board, Washington DC. Available at
      $File/SAB-05-003 unsigned.pdf (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 2008a. Waste and Materials-Flow Benchmark
      Sector Report: Beneficial Use of Secondary Materials - Foundry Sand. Office of Solid
      Waste. February.
U.S. EPA (Environmental Protection Agency). 2008b. Child-Specific Exposure Factors
      Handbook. EPA/600/R-06-096F. U.S.  Environmental Protection Agency, Office  of
      Research and Development, National Center for Environmental Assessment,
      Washington, DC.  October. Available at
      http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid= 199243 (accessed 30 June 2014).
U.S. EPA (Environmental Protection Agency). 2009a. Integrated Risk Information System
      (IRIS). National Center for Environmental Assessment, Office of Research and
      Development, Washington, DC. Available at http://www.epa.gov/iris/ (accessed 30 June
U.S. EPA (Environmental Protection Agency). 2009b. Human and Ecological Risk Assessment
      of Coal Combustion Wastes, Draft. Office of Resource Conservation and Recovery.
      EPA530-D-09-001, August.
U.S. EPA (Environmental Protection Agency). 2011. Exposure Factors Handbook: 2011
      Edition. EPA/600/R-090/052F. U.S. Environmental Protection Agency, Office of
      Research and Development, Washington, DC. September. Available at
      http://cfpub.epa.gov/ncea/risk/recordisplay.cfm?deid=236252 (accessed 30 June 2014).

Responses to Review Comments                                Appendix A
                      Appendix A:

    IWEM modeling review, alternate model search, and
   recommendations for evaluating the SFS home garden
             scenario groundwater pathway


Responses to Review Comments                                                Appendix A

Appendix A) IWEM modeling review, alternate model search,
and recommendation for evaluating the SFS home garden
scenario  groundwater pathway

A1.0 Background

The Peer Review Draft Spent Foundry Sands (SFS) evaluation used EPA's Industrial Waste
Evaluation Model (IWEM) to estimate human exposure to contaminants leached from a SFS-
amended home garden via contamination of groundwater in a nearby drinking water well.4
IWEM was chosen because it is a peer-reviewed, publically available model that has successfully
supported regulatory decisions, and it can model a waste management scenario that is similar to
the SFS home garden scenario. Specifically, IWEM's land application unit (LAU) waste
management scenario had been run as the waste management scenario most  similar to the SFS
home garden scenario.

External peer-review comments on the draft SFS evaluation led EPA test the IWEM-generated
SFS home garden scenario receptor well concentration estimates. This subsequent testing raised
the concern that the choice of the LAU waste management scenario, and input parameter values
used when modeling the LAU scenario (e.g., waste management unit operating life), may have
underestimated closest well concentrations for the SFS home garden scenario. It was also
possible that IWEM was not  the most appropriate model: another model may more accurately
estimate closest groundwater exposures in the SFS home garden scenario.

A2.0 Purpose and Objectives

To address these concerns, EPA first conducted a thorough review of the groundwater exposure
modeling performed for the Peer Review Draft SFS evaluation, to fully understand the
implications of input parameter choices used when implementing the LAU waste management
scenario. Second,  EPA compared the various waste management scenarios available within
IWEM (i.e., in addition to LAUs, IWEM is able to model landfills, surface impoundments, and
waste piles), including input  parameter choices, to identify which scenario and input parameter
choices would most accurately estimate groundwater exposures for the SFS home garden
scenario. Third, a  search was conducted to identify whether there are any peer-reviewed and
publically available groundwater models, of good standing in the regulatory  arena, which could
be used to more accurately represent the SFS home garden scenario groundwater exposure
pathway in a national-scale assessment.

This appendix presents the evaluation findings and recommends the most appropriate model and
approach to evaluate the SFS home garden groundwater pathway. Section A3 discusses in detail
the review of the Peer Review Draft SFS IWEM modeling, and comparison of IWEM waste
4 IWEM supports two levels of analysis: Tier 1 (a screening-level analysis using default data values based on
 national distributions for many parameters) and Tier 2 (a site-specific analysis based on location-adjusted values
 for the most sensitive waste- and site-specific parameters). The SFS evaluation was based on Tier 2, using three
 locations to represent variability in meteorological conditions. Thus, "IWEM" in this memorandum refers to that
 Tier 2 analysis.

                                                                       Page 1 of 17

Responses to Review Comments                                                   Appendix A

management scenario options. Section A4 discusses the models identified and considered as
alternatives to IWEM. Based on results of the review, scenario comparison, alternative model
search, IWEM remains the preferred model for supporting the SFS assessment. Section A5
provides the recommendations on how best to estimate receptor well concentrations for the SFS
home garden scenario.

A3.0 Review of SFS home  gardener scenario modeling, and
       IWEM waste management scenario comparison

The numerical engine for IWEM is EPA's Composite Model for Leachate Migration with
Transformation  Products (EPACMTP). The review of the Draft SFS Risk Assessment
groundwater exposure modeling included analyzing how IWEM employs EPACMTP to model
the LAU scenario. This analysis uncovered a number of differences between the SFS home
garden scenario and IWEM's LAU scenario. To develop a better understanding of key
underlying IWEM modeling assumptions and limitations, EPA investigated a number of critical
modeling choices (i.e., inputs and options available in IWEM) and issues identified by EPA,

   1.  Well Under Garden Scenario: The residential receptor well closest to the SFS home garden
       could arguably exist under the home garden, which IWEM does not allow. The SFS evaluation
       initially placed the well 1 meter from the edge of the garden, the smallest distance to the receptor
       well that  IWEM allows.

   2.  Square Pulse and Conservation of Mass: The Draft SFS risk assessment assumed that the SFS
       would be applied in a single application and remain in the home garden (with SFS constituents
       available for release), so that contaminant mass would be released until all of the available mass
       had been depleted, and the concentration would presumably lessen over time. In contrast, the
       LAU implementation in IWEM/EPACMTP assumes that contaminant mass is applied regularly
       to an area at a constant rate over a finite time period, resulting in a leaching pattern that is
       constant over the time period and then stops, reflecting the end of land application and depletion
       or removal of the source (i.e., a "square pulse" source). The PvVEM LAU implementation defines
       the end of this square pulse through a finite "operational lifespan" (with a default of 40 and a
       maximum of 200 years). In short, the PvVEM implementation of EPACMTP models all LAUs as
       temporary (i.e. "pulse") sources, with a constant leaching concentration during the "operational
       lifespan," and a leachate concentration of zero for all modeled years after that time.

   3.  Operational Lifespan and Timestep: EPA wanted to better understand how operational life and
       the initial time step for "testing" receptor well concentrations interact when modeling a pulse
       source. Specifically, does EPACMTP use the operational lifespan as an initial time step to choose
       when to "test" receptor well concentrations? When a user specifies the LAU's operational life,
       does PWEM force EPACMTP to use the default 40-year value as the initial time step regardless of
       the user's specified value? In this case, EPA wanted to ensure that the peak groundwater
       concentration does not pass the receptor well before the well is "tested," especially for receptor
       wells close to the LAU.

   4.  1-Year Operational Lifespan: The SFS evaluation assumed a single application of
       manufactured soil amended with SFS while constructing a home garden, which was represented
       in PvVEM by assuming an LAU operational lifespan of 1 year. EPA wanted more information on
       the impacts of this assumption on the model results.
                                                                           Page 2 of 17

Responses to Review Comments                                                   Appendix A

Key components and implications of these issues are addressed in the following sections: the
well under the garden scenario (Section A3.1); the use of a square pulse and conservation of
mass (Section A3.2); LAU operational life and initial EPACMTP time step (Section A3.3); and
how receptor well  distance and LAU operational life assumptions interact to influence model
results (Section A3.4). The analyses presented in these subsections were designed to test whether
any of the IWEM assumptions/limitations used in the Draft SFS Risk Assessment modeling, or
combinations thereof, could underestimate SFS home gardener receptor well concentrations.

A3.1  Residential Well Under the Garden

The residential well was assumed to be 1 meter from the edge of the unit rather than directly
under it. IWEM/EPACMTP does not permit the well to be located under the source, so it is not
possible to test this assumption using EPACMTP. However, it is unlikely that this limitation
would be associated with a significant difference in predicted well concentrations. Specifically, a
well located under the center of a garden would be exposed to less contamination than one at the
downgradient edge, because the well in the center would receive contaminated infiltration from
only the upgradient portion of the garden instead of all of it. Also, because the well depth was
varied during the IWEM SFS runs, there is a good possibility that the well would be exposed to
more clean water under the garden at greater well  depths than it would at the short (1 m) distance
from the edge of the unit, because the plume would not have had time to mix into the aquifer.
Therefore, locating the well 1 meter from the edge of the garden is a conservative assumption
when compared to locating the well directly under the garden.

A3.2  Square Pulse and Conservation of Mass

One concern with the SFS garden scenario is the apparent disparity in how mass  is applied and
released when comparing the SFS scenario with the conceptual model of how an LAU is
implemented in IWEM/EPACMTP. In the SFS home garden scenario, it is reasonable  to expect
leachate concentrations to decrease over time after a single application. By contrast, the
conceptual model of the LAU in IWEM/EPACMTP uses a "square pulse" LAU source term, in
which the leachate concentration remains relatively constant until the source is depleted (when it
drops to zero). However, the observed leaching behavior of arsenic, a solubility-controlled
constituent, is in many waste disposal environments more consistent with the IWEM "square
pulse" conceptual model than the single-application-followed-by-decay scenario posited for the
home garden. Furthermore, the "square pulse" LAU source term is a conservative simplifying
assumption with respect to establishing screening  level criteria for mass loading. Mass can be
conserved under the "square pulse" scenario by considering the available mass and infiltration
rate to limit the pulse length.

In addition, the nonlinear sorption transport module in EPACMTP, used for metal simulations in
the unsaturated zone, applies a similar simplifying square pulse assumption to all source term
conceptual models as a trade-off for computational efficiency. Nonlinear transport is a complex
computational problem which is compounded when conducting Monte Carlo simulations; a
square pulse is a necessary and appropriate simplification to maintain reasonable computational
times. Regardless of the characteristic shape of the leachate concentration  profile over  time,
IWEM/EPACMTP would represent the correct total mass loading with an equivalent square
                                                                           Page 3 of 17

Responses to Review Comments                                                   Appendix A

pulse. Therefore, the square-pulse assumption is considered a reasonable simplification that does
not alter the results significantly.

A3.3 Operational Lifetime and Initial EPACMTP Time Step

IWEM/EPACMTP does not use the operational lifetime of an LAU (or any other WMU) to
determine the time stepping strategy for identifying the peak or average concentration at a
receptor well during flow and transport simulations. The adaptive aspect of EPACMTP's time
stepping strategy is based primarily on the transport simulation results from the unsaturated zone
and the fate and transport characteristics of the contaminant. Specifically, the unsaturated zone
transport modules generate a time series of concentrations at the water table. That time series
captures both the arrival time of the leachate "front" at the water table and the time at which
leachate no longer arrives. Those two times are used to define the maximum duration of the
contaminant pulse. The contaminant pulse's onset and ending times, in conjunction with
knowledge about sorption and decay of the contaminant in the saturated zone, are used to predict
when the contaminant will likely arrive at the receptor well and when the plume is likely to pass
by that location. The prediction of arrival time is tempered with a safety factor to reduce the risk
of predicting an arrival time that is too late, thus missing the peak. IWEM/EPACMTP discretizes
the time interval until the predicted arrival time and steps through that time interval linearly to
identify the peak concentration. Thus, the likelihood of IWEM missing the peak concentration
due to an inappropriately long time step is small, and is unrelated to operational lifetime.

A3.4 Well Distance and Operating Life Assumptions

The remaining issues revolve around the operating life (1 year) and well distance (1 m) assumed
for the SFS home garden scenario and how these interact. These variables are connected, and
because both of these values are small, they combine to generate a 90th percentile result that is
low when compared to results obtained with a higher well distance or a more typical  "operating
life" assumption (like the default 40 years). For these reasons, these variables were treated
together in this analysis, which focused first on well distance at the 1-year operating life and then
on the effects of increasing the operating life. Implications for modeling EPA's SFS scenario are
then discussed.

IWEM was used to model several well-distance permutations of the SFS home gardener
scenario. Results were analyzed graphically and numerically by visualizing and summarizing the
peak concentration values for each run. These values are generated as standard outputs during a
Tier 2 analysis by IWEM in the *.SAT output file that is generated during each model run.
Original *.SAT files were saved along with their associated IWEM run/project files (*.wem and
*.mdb). Analyses were performed using the Python programming language and widely used
open-source scientific computing libraries.

A3.4.1 Characterize Modeled Exposure Results at Various Well Distances for As+3

IWEM was initially run 23 times, modeling trivalent arsenic (As+3, the primary risk driver for the
SFS risk analysis), and varying only the well distance for each run. A list of assumptions and
well distances used in this modeling is provided in Table A-l. The resulting 90th percentile peak
concentrations reported for the 23 runs are displayed in Figure A-l and Table A-2.
                                                                           Page 4 of 17

Responses to Review Comments
Appendix A
Results demonstrate a rapid increase in concentration with distance in wells less than 5m from
the source, with the highest peak concentration occurring at a well distance of 30 m. The 30 m
peak concentration is more than 5 orders of magnitude higher than the peak concentration at the
1 m well distance. These results are somewhat counterintuitive, as one would normally expect
higher groundwater concentrations closer to the source.

These unexpected results were investigated further by evaluating the well concentrations at
different screen depths provided in the IWEM outputs.
           Table A-l. Assumptions and Variables Used in Initial SFS Simulations
Parameter Name
Source type
LAU operational life
Well distances (m)
Subsurface environment
Soil type
Leachate concentration (mg/L)
Number of Monte Carlo
Land application unit (LAU)
1 year
1-15 (inclusive, at 1-m intervals), 30, 50, 75, 100, 150, 200
Unknown (sets default values for groundwater pH, depth to
aquifer hydraulic conductivity, regional hydraulic gradient,
, 250, 300
water table,
and aquifer
Seattle, WA
Arsenic III, CAS ID 7440-38-2
0.0156 (95th percentile of SFS leachate measurements)
                        . ie-4  IWEMv2 SFS Scenario Exposure Results at Various Well Distances
                                       100      150     200
                                        User-defined well distance (m)
             Figure A-1. 90th Percentile Peak Well Concentration with Well Distance:
                      Initial SFS Simulation, Home Garden Scenario, As*3
                                                                               Page 5 of 17

Responses to Review Comments
Appendix A
    Table A-2. 90th Percentile Peak Trivalent Arsenic Concentration with Well Distance:
                   Initial SFS Simulation, Home Garden Scenario, As+3
Well Distance
90th Percentile Peak
Concentration (mg/L)
Well Distance
90th Percentile Peak
Concentration (mg/L)
Figure A-2 shows eight curves: four depict various percentile statistics of the peak concentration
based on the model results for all well depths, while the other four depict the same percentile
statistics from the results with well depth constrained to a range of 0 to 4 meters. The key
observations are that (1) the 90th percentile values at well distances less than 30 m are
significantly lower than the peak concentration values at greater well distances when all depths
are considered and (2) this effect does not occur for percentiles of 95 and above or for 90th
percentile values constrained to 0-4 m well depths. This suggests that the extremely low
concentrations for the lowest well distances occur because the plume has not mixed very deeply
within the aquifer, and the random well depth within the aquifer is sampling clean groundwater
below the contaminant plume. When only shallow well depths are considered (i.e., 0-4 m), the
results follow a continuously decreasing trend with well distance  similar to what would be
expected in a typical groundwater plume.
                                                                             Page 6 of 17

Responses to Review Comments
                                                                    Appendix A
Responses to Review Comments
Appendix A
To investigate the sensitivity of the model results to operating life, IWEM also modeled As+3 at
operating lifetimes close to the 200-year maximum and the 40-year default for this parameter,
which are representative of the As+3 depletion time, which is  195 years at an initial As+3 whole
waste concentration close to the 95th percentile and 44 years at an initial concentration close to
the 50th percentile, based on percentiles reported for SFS wastes.5 Results are shown in Table A-

The home garden was also run as a landfill sized to be equivalent to the SFS LAU (1-acre, 0.2-m
thick layer of 50% SFS and 50% soil). This was because IWEM/EPACMTP models landfills as  a
depleting source. Whereas the modeling duration of the LAU model is governed by the LAU's
operational life, IWEM/EPACMTP depletes the waste concentration in a landfill, extending the
modeled duration until the mass of constituent is exhausted, yielding an essentially steady state
transport result. Otherwise the EPACMTP landfill operates very similarly to the LAU, and is
subject to the square leachate pulse assumption for nonlinear  sorption  of metal constituents.
Table A-3 also includes landfill modeling results.

            Table A-3. Effect of Well Distance, Operating Life, and WMU Type
                on As+3 Receptor Well Concentrations for a 1-Scre WMU
Well distance
90th Percentile As+3 Peak Concentration (mg/L)
1-year LAU
40-year LAU
195-year LAU
           Leachate Concentration (Cl) = 0.0156 mg/L, 95th percentile leachate concentration.

Table A-3 shows that, for As+3 in  SFS, there is essentially no difference in results between the
40- and 195-year LAUs and the landfill, with the highest concentrations occurring at the 1 m
well, and that all results run for the longer "operating life" are more than an order of magnitude
higher than the Tapwater screening level.

Figure A-4 shows the various percentiles of concentration for various depths and operating lives.
The results for the 1-year LAU scenario show the anomalously low values for the closer well
distances as discussed above.

Figure A-5 plots the IWEM output data to show concentration by depth and distance. The 1-year
LAU results show significantly lower (about 2 orders of magnitude) values than either the 40-
year or 195-year LAU runs, which can be attributed to the wastes not being left in place long
enough to fully develop the contaminant plume and reach peak groundwater concentrations.
Figure A-5 also shows that the 40-yr and 195-year plumes are very similar for As+3.
5 If a 50:50 mixture of soil and SFS is considered, the depletion times for 95th and 50th percentile waste
 concentrations would be 97 and 22 years respectively. Note that IWEM/EPACMTP does not consider whole waste
 concentrations in any of its WMU release calculations.
                                                                              Page 8 of 17

Responses to Review Comments
Appendix A
                        IWEMv2 5F5 Scenario Exposure Results - LAU lyr vs. 40yr (Aslll)
                                                        lyr 90th pctile (all depths)
                                                        lyr 95th pctile (all depths)
                                                        lyr 99th pctile (all depths)
                                                        lyr 90th pctile (0-4m depths)
                                                        lyr 95th pctile (0-4m depths)
                                                        lyr 99th pctile (0-4m depths)
                                                        40yr 90th pctile (all depths)
                                                        40yr 95th pctile (all depths)
                                                        40yr 99th pctile (all depths)
                                                        SFS groundwater screening level
                                            20          30
                                       User-defined well distance (m)
 Figure A-4. Comparison of 1-year and 40-year operating lifetimes for the LAU SFS home
  garden scenario for As+3 results. (Green line = 4.5*10~4 mg/L Tapwater screening level.)
                                     IWEMV2 SFS Plume lyr LAU (A
                                         User-defined well distance (m)
                                                               IWEMVZ SFS Plume - 195yr LAU (Aslll)
                User-defined well distance (m)
                                                                   User-defined well distance (m)
 Figure A-5. As+3 concentration plots (by well distance and depth) for 1-year (top), 40-year
 (middle), and 195-year LAU (bottom) simulations (with all other parameters set the same).
 Note that concentration scale is different (lower) for 1-year simulations to show "plume".
                                                                                     Page 9 of 17

Responses to Review Comments
Appendix A
A3.4.3 Characterize Modeled Exposure Results for Different Operating Lifetimes and Well
       Distances for Other Metals
In addition to As+3, the SFS groundwater risk assessment modeled four metals (antimony,
beryllium, cadmium, and lead) using one-half their respective detection limits. These
constituents were modeled using the same LAU model conditions used for the original modeling
scenario conditions (1-year operating life and 1-m well distance), as well as the same range of
well distances (1m, 15m, 30m, 100m) and operating lifetimes (1-year, 40-year, 200-year) as
investigated for As+3. These constituents were also modeled using the IWEM/EPACMTP landfill
model with the depleting source scenario. Results are shown below.

Table A-4 provides results for antimony, which behaved similarly to arsenic, with the 1-year
results being about an  order of magnitude below the rest, and the 40-year, 200-year, and landfill
results being essentially equivalent.

          Table A-4. Effect of Well Distance, Operating Life, and WMU Type on
                Antimony Receptor Well Concentrations for 1-Acre WMU
Well distance
90th Percentile Antimony Peak Concentration (mg/L)
(0.2 m deep)
               Leachate concentration (0.02 mg/L) equal to one-half the detection limit.

Table A-5 provides results for beryllium, for which the 1-year 90th percentile concentrations
were zero at all well distances, suggesting that this was a strongly sorbing constituent with a
large attenuation in concentration in the vadose zones. This is also reflected in the approximate
order of magnitude difference between the 200-year LAU and the landfill results with the landfill
giving the higher groundwater concentration.

     Table A-5. Effect of Well Distance, Operating Life, and WMU Type on Beryllium
                     Receptor Well Concentrations for 1-Acre WMU
Well distance
90th Percentile Beryllium Peak Concentration (mg/L)
(0.2 m deep)
               Leachate concentration (0.01 mg/L) equal to one-half the detection limit.
                                                                           Page 10 of 17

Responses to Review Comments
Appendix A
The results for cadmium (Table A-6) and lead (Table A-7) show a behavior very similar to
arsenic and antimony, with very little difference between the 40-year LAU, 200-year LAU, and
landfill results but with lower concentrations for the 1-year LAU results, particularly for the 1 m
well distance, and especially for cadmium.

          Table A-6. Effect of Well Distance, Operating Life, and WMU Type on
                Cadmium Receptor Well Concentrations for 1-Acre WMU
Well distance
90th Percentile Cadmium Peak Concentration (mg/L)
1 -year LAU
(0.2 m deep)
               Leachate concentration (0.005 mg/L) equal to one-half the detection limit.
          Table A-7. Effect of Well Distance, Operating Life, and WMU Type on
                  Lead Receptor Well Concentrations for 1-Acre WMU
Well distance
90th Percentile Lead Peak Concentration (mg/L)
(0.2 m deep)
               Leachate concentration (0.055 mg/L) equal to one-half the detection limit.

A3.4.4 Conclusions Regarding Operating Life and Well Distance
Based on the analyses above, it is evident that concentration does not behave as expected for a 1-
year "operating" life, with maximum peak concentrations 30 m from the garden in this scenario.
However, for a more reasonable operating life (e.g., 40 years), maximum peak concentrations do
occur at the closest well distance as expected.
                                                                           Page 11 of 17

Responses to Review Comments
                                    Appendix A
A4.0 Alternative Model Evaluation
A search was conducted to identify models
that could potentially serve as suitable
alternatives to IWEM for supporting the
evaluation of the SFS home garden
groundwater pathway. Ideally, the models
would need to be capable of (1) estimating
leachate concentration and (2) simulating the
transport of metal constituents to
groundwater drinking wells. The only
models that were considered as alternatives
to IWEM were peer-reviewed, publically
available, and used to support regulatory

Table A-8 summarizes the capabilities and
characteristics of the models identified for
consideration. Based on a model
comparison, IWEM remains the preferred
model for supporting the SFS evaluation;
specifically, the model's capabilities (e.g.,
probabilistic modeling6, nonlinear metal
sorption isotherms—see text box), and
history (extensive peer-review record, long
history of supporting regulatory decisions)
were clearly superior to the alternatives. For example, EPACMTP, the numerical engine for
IWEM, was also evaluated separately as  an alternative to IWEM. However, given the resources
needed to parameterize and execute EPACMTP, IWEM provides an efficient, cost-effective, and
scientifically defensible screening tool; EPACMTP is recommended for refined modeling for
constituents that fail an IWEM analysis. In addition, none of the other identified alternative
models offer the probabilistic (i.e., Monte Carlo) modeling framework used in IWEM, and only
one model combination, the SEasonal SOIL (SESOIL) compartment model coupled with the
groundwater model Analytical Transient 1-, 2-, and 3-Dimensional Simulation of Waste
Transport in the Aquifer System (AT123D), includes nonlinear isotherms and is associated with
State and EPA remediation efforts. However, the criterion of being publically available was not
fully met by this combination of models: although the source code for SESOIL and AT123D is
free, the user interface is not, and is available only through commercial vendors.
          Nonlinear Sorption Isotherms
A nonlinear sorption isotherm is an expression of the
equilibrium relationship between the sorbed
concentration of a metal (or other constituent) and the
aqueous concentration for a representative set of
subsurface system conditions. Nonlinear sorption
isotherms  are important when modeling metals because
metal sorption coefficients (Kds), which influence metal
fate and transport, are significantly affected by metal
concentration in the aqueous phase.  In general,  metal
mobility tends to be higher (and thus, Kds lower) as
leachate concentrations increase. Therefore, as
leachate concentrations decrease during unsaturated
zone (soil) transport, metal mobility also tends to
decrease (and Kds tend to increase).

The use of nonlinear metal sorption isotherms enables
EPACMTP to model nonlinear behavior in the
unsaturated zone module for a wide array of subsurface
conditions. For the  sorption isotherms used in IWEM
EPACMTP, these conditions are defined by parameters
that include the pH  of the aqueous system,
concentrations of adsorbents, natural organic matter,
anthropogenic organic acids, and other characteristics
appropriate for particular waste streams. IWEM includes
ensembles of nonlinear isotherms that have been
compiled for hundreds of combinations of these
parameters for more than 20 metals for selection and
use in the  probabilistic modeling framework.
6 With limited input of setting and source characteristics, IWEM provides the user with user-friendly access to
 EPACMTP, which contains nationwide aquifer, soil, rainfall/infiltration, and metal sorption data sets, and a
 probabilistic modeling framework capable of representing variability and uncertainty in model inputs for a specific
 site area, region, or the entire United States.
                                                                                Page 12 of 17

Responses to Review Comments
Appendix A
                                    Table A-8. Comparison of Alternative Models

EPACMTP is a full-featured
groundwater flow and transport
model with probabilistic
modeling capabilities with a long
history of supporting EPA
regulatory development for the
management and disposal of
hazardous wastes. Designed to
simulate subsurface fate and
transport of contaminants
released from land disposal sites.
Predicts groundwater exposures
in domestic drinking water
receptor wells Simulations are
performed using probabilistic
input specifications based on
nationwide data.

Parameterizing and executing EPACMTP can be
challenging, but the model includes resident
databases that allow the model to be used for
nationwide assessments using Monte Carlo
simulation techniques; it is not intended for site-
specific applications. The Monte Carlo module
of EPACMTP allows you to take into account
the effect of parameter variability on predicted
ground-water concentrations.

dispersion, linear
or nonlinear
sorption, and
reactions. In cases
where degradation
of a waste
constituent yields
daughter products
that are of concern,
accounts for
formation and
transport of up to
six different
daughter products.


                                                                                                      Page 13 of 17

Responses to Review Comments
                                                                        Appendix A
 (Pesticide Root Zone
 Model for Ground
EPA's Office of Pesticide
Programs detailed, numerical
solution model which is typically
used to model pesticide leaching.
The model can be used to
support screening (Tier I) and
refined (Tier II) drinking water
Model represents vulnerable private drinking
water wells in the vicinity of agricultural
environments. The saturated zone of the
conceptual model is a shallow unconfmed aquifer
with a water table depth that corresponds to the
scenario location. Well-screen extends from
aquifer surface to 1 m below surface, but this
length is adjustable. Provides six standard
scenarios that represent regions known to have
vulnerable GW supplies. For Tier I assessments,
it is recommended that simulations be run with
all six scenarios. A simulation can be run for up
to 100 years. Tier II refinement options include
(1) development of representative scenario (e.g.,
soil type and characteristics, weather data, depth
to aquifer); (2) identify pesticide fate parameters
not considered in the Tier 1 Simulations
(subsurface degradation and subsurface
sorption);  (3) change application reoccurrence;
(4) considerations of well setbacks; (5) explore
different exposure durations.
Degradation and
linear sorption
 Concentration /n
 G^Ound Water)0
EPA OPP's very simple,
conservative screening model
used to develop an estimate of
likely ground-water
concentrations if a pesticide is
used at the maximum allowable
rate in areas with ground water
exceptionally vulnerable to
A simple user interface with 6 inputs. Intended
for estimating conservative or high-end exposure
values because the model is based on ground-
water monitoring studies which were conducted
by applying pesticides at maximum allowed rates
and frequency to vulnerable sites. Does not have
the capability to consider variability in leaching
potential of different soils, weather (including
rainfall), cumulative yearly applications or depth
to aquifer.
Degradation and
linear adsorption
                                                                                                                                  Page 14 of 17

Responses to Review Comments
Appendix A

(SEasonal SOIL
compartment model)
/ (Analytical
Transient 1-, 2-, and
3 -Dimensional
Simulation of Waste
Transport in the
Aquifer System)d

VLEACH (Vadose
zone LEACHing
model), Version

SESOIL is a one -dimensional
soil compartment model.
Downward transport through soil
column. Has been linked to the
ground water transport model
(AT123D). Has become fairly
well established and accepted by
several state agencies and the
U.S. EPA for calculating
remediation standards.
Numerical solution and
screening model with a one-
dimensional, finite difference
model for making preliminary
assessments of the effects on
ground water from the leaching
of volatile, sorbed contaminants
through the vadose zone. Not
intended for modeling metal

Both SESOIL and AT123D source codes are
available free of charge from the U.S. EPA but
they lack a user interface. Commercially
available interfaces are available.

Model execution is initiated from a DOS
command line and inputs are read from manually
created, structured text input files.

transport, and
metal nonlinear
isotherms for

sorption, vapor-
phase diffusion,
and three-phase
Linear isotherms
describe the
partitioning of the
pollutant between
the liquid, vapor
and soil phases




                                                                                                          Page 15 of 17

Responses to Review Comments                                                  Appendix A
A5.0 Recommendations

Based on the results of the model review and detailed analysis of the IWEM/EPACMTP results
for the SFS home garden scenario, the following recommendations are made ( and instituted in
the Final SFS risk assessment) to support SFS home gardener groundwater pathway screening:
   •   Continue to use IWEM. IWEM meets or exceeds all of the criteria specified for this
       review, and contains features (like the probabilistic modeling framework and nonlinear
       sorption isotherms for metals) that are not available in most of the other identified
   •   Change the modeled operating life. Set operating life to the default (40-year) and
       maximum (200-year) values to allow sufficient time for the plume to develop and for the
       peak concentrations to be achieved at the receptor well. The 1-year "operating life" as
       implemented in the Peer Review Draft SFS modeling effort removes the SFS-amended
       soil after 1 year, which does not reflect how most people manage their gardens (i.e.,
       garden soils are likely always left in place). Because removing the amended soil
       effectively stops the leaching process after 1 year, this significantly underestimates the
       well concentrations because it does not allow sufficient time for the contaminant plume
       to develop before the simulation is stopped.
   •   Continue to use aim receptor well distance. This well distance is acceptable and
       reasonable considering the SFS home garden exposure scenario, and gives the highest
       concentrations when compared to other well distances as long as a sufficient and
       reasonable modeling time (i.e., "operating life") is  used.
   •   Use 10,000 flow and transport simulations. 5,000 simulations is not sufficient to ensure
       stability in the tails of the resulting distribution of receptor  well concentrations.
Appendix A: References
N.J. Department of Environment Protection (DEP). 2008a. Guidance Document Using SESOIL
       Transport Model to Assess the Impact to Ground Water Pathway. Available online at
       http://www.nj.gov/dep/srp/guidance/rs/sesoil.pdf (accessed 30 June 2014).

N.J. Department of Environment Protection (DEP). 2008b. Guidance for Using the SESOIL and
       AT 12 3D Models to Develop Site Specific Impact to Ground Water Soil Remediation
       Standards. Available online at
       http://www.nj.gov/dep/srp/guidance/rs/atl23d guidance.pdf (accessed 30 June 2014).

U.S. EPA (Environmental Protection Agency). 1997a. EPA 's Composite Model for Leachate
       Migration with Transformation Products. EPACMTP: User's Guide. Office of Solid
       Waste, Washington, DC. Available online at
       http://www.epa.gov/osw/nonhaz/industrial/tools/cmtp/userguid.pdf (accessed 30 June

U.S. EPA (Environmental Protection Agency). 1997b. VLEACHA One-Dimensional Finite
       Difference Vadose Zone Leaching Model, Version 2.2a. Office of Research and
       Development, Ada, Oklahoma. Available online at
       http://www.epa.gov/nrmrl/gwerd/download/vleach.pdf (accessed 30 June 2014).
                                                                         Page 16 of 17

Responses to Review Comments                                                  Appendix A

U.S. EPA (Environmental Protection Agency). 2003. SCI-GROW Version 2.3. Office of
       Pesticide Programs, Washington, DC. Available online at
       http://www.epa.gov/oppefedl/models/water/scigrow description.htm (accessed 30 June
U.S. EPA (Environmental Protection Agency). 2012. Guidance for Using PRZM-GW in
       Drinking Water Exposure Assessments. Office of Pesticide Programs, Washington, DC.
       Available online at http://www.epa.gov/oppefed 1 /models/water/przm gw/
       wqtt_przm_gw_guidance.pdf (accessed 30 June 2014).
                                                                         Page 17 of 17