EPA’s Response to Peer Review Comments on the Formaldehyde Indoor Air
Model – Pressed Wood Products (FIAM-pwp)

U.S. Environmental Protection Agency

Office of Pollution Prevention and Toxics

Exposure Assessment Branch

1200 Pennsylvania Avenue, N.W.

Washington, D.C.  20460

 July 2012

EPA acknowledges the analytical and draft preparation support of Versar,
Inc. of Springfield, Virginia in the preparation of this report,
provided under Contract No. EP-W-10-005.

Table of Contents

  TOC \o "1-3" \h \z \u    HYPERLINK \l "_Toc320006449"  1 INTRODUCTION	
 PAGEREF _Toc320006449 \h  1  

  HYPERLINK \l "_Toc320006450"  1.1	Background	  PAGEREF _Toc320006450
\h  1  

  HYPERLINK \l "_Toc320006451"  1.2	This Report	  PAGEREF _Toc320006451
\h  2  

  HYPERLINK \l "_Toc320006452"  2 SUMMARY OF PEER REVIEW COMMENTS AND
EPA RESPONSES, ORGANIZED BY CHARGE QUESTION	  PAGEREF _Toc320006452 \h 
3  

  HYPERLINK \l "_Toc320006453"  2.1	Part 1 - Algorithms and Exposure
Assumptions	  PAGEREF _Toc320006453 \h  3  

  HYPERLINK \l "_Toc320006454"  2.1.1 	Comment on the Adequacy of the
Mathematical Equations and the Level of Confidence in Indoor Air
Concentrations and Exposure Estimates Provided by the Model	  PAGEREF
_Toc320006454 \h  3  

  HYPERLINK \l "_Toc320006455"  2.1.2 	Comment on the Default Scenarios
and Assumptions Used by the Model	  PAGEREF _Toc320006455 \h  8  

  HYPERLINK \l "_Toc320006456"  2.1.3 	Comment on the Data Sources	 
PAGEREF _Toc320006456 \h  9  

  HYPERLINK \l "_Toc320006457"  2.1.4 	Comment on Whether the Data
Sources Provide Unused Information That Would Assist in Quantifying
Uncertainty or Variability for Subsequent Model Use	  PAGEREF
_Toc320006457 \h  9  

  HYPERLINK \l "_Toc320006458"  2.1.5 	Comment on Whether the Model
Input Variables and the Estimates Assigned to Them Appropriately Reflect
Variability	  PAGEREF _Toc320006458 \h  10  

  HYPERLINK \l "_Toc320006459"  2.1.6 	Comment on the Adequacy of the
Documentation Supporting the Overall Plausibility of the Exposure
Estimates Generated by the Exposure Model	  PAGEREF _Toc320006459 \h  10
 

  HYPERLINK \l "_Toc320006460"  2.1.7 	Provide Any Other Types of Data
to Assess the Reliability of the Overall Exposure Model or Specific
Model Elements	  PAGEREF _Toc320006460 \h  11  

  HYPERLINK \l "_Toc320006461"  2.1.8 	General Comments on Algorithms
and Exposure Assumptions	  PAGEREF _Toc320006461 \h  11  

  HYPERLINK \l "_Toc320006462"  2.2	Part 2 - User Interface	  PAGEREF
_Toc320006462 \h  12  

  HYPERLINK \l "_Toc320006463"  2.2.1 	Provide Comments on the Ease of
Use of the FIAM-PWP Model	  PAGEREF _Toc320006463 \h  12  

  HYPERLINK \l "_Toc320006464"  2.2.2 	Comment on the Adequacy of the
Model Help Screens	  PAGEREF _Toc320006464 \h  12  

  HYPERLINK \l "_Toc320006465"  2.2.3 	General Comments on User
Interface	  PAGEREF _Toc320006465 \h  12  

  HYPERLINK \l "_Toc320006466"  2.3	Part 3 – Documentation	  PAGEREF
_Toc320006466 \h  13  

  HYPERLINK \l "_Toc320006467"  2.3.1 	Comment on Whether the References
Have Been Accurately Identified	  PAGEREF _Toc320006467 \h  13  

  HYPERLINK \l "_Toc320006468"  2.3.2 	Provide Comments on the Model
Documentation (Does It Present the Model Construct, Selected Model
Inputs, and Model Results in a Clear, Complete and Useful Manner?)	 
PAGEREF _Toc320006468 \h  14  

  HYPERLINK \l "_Toc320006469"  2.3.3 	Provide Comments on Whether the
Documentation for the FIAM - PWP Makes Clear the Sources and Anticipated
Magnitudes of Variability and Uncertainty in Model Results	  PAGEREF
_Toc320006469 \h  15  

  HYPERLINK \l "_Toc320006470"  2.3.4 	Comment on Whether the Model
Algorithms are Consistent with the Descriptions in the Model
Documentation	  PAGEREF _Toc320006470 \h  15  

  HYPERLINK \l "_Toc320006471"  2.3.5 	General Comments on Documentation
  PAGEREF _Toc320006471 \h  15  

  HYPERLINK \l "_Toc320006472"  APPENDIX A – DETAILED REVIEWER
COMMENTS AND    HYPERLINK \l "_Toc320006473"  EPA RESPONSES:       Part
1 – Algorithms and Exposure Assumptions 

  HYPERLINK \l "_Toc320006474"  2.1.1 	Comment on the Adequacy of the
Mathematical Equations and the Level of Confidence in Indoor Air
Concentrations and Exposure Estimates Provided by the Model	A-  PAGEREF
_Toc320006474 \h  1  

  HYPERLINK \l "_Toc320006475"  2.1.2 	Comment on the Default Scenarios
and Assumptions Used by the Model	A-  PAGEREF _Toc320006475 \h  11  

  HYPERLINK \l "_Toc320006476"  2.1.3 	Comment on the Data Sources	A- 
PAGEREF _Toc320006476 \h  12  

  HYPERLINK \l "_Toc320006477"  2.1.4 	Comment on Whether the Data
Sources Provide Unused Information That Would Assist in Quantifying
Uncertainty or Variability for Subsequent Model Use	A-  PAGEREF
_Toc320006477 \h  13  

  HYPERLINK \l "_Toc320006478"  2.1.5 	Comment on Whether the Model
Input Variables and the Estimates Assigned to Them Appropriately Reflect
Variability	A-  PAGEREF _Toc320006478 \h  14  

  HYPERLINK \l "_Toc320006479"  2.1.6 	Comment on the Adequacy of the
Documentation Supporting the Overall Plausibility of the Exposure
Estimates Generated by the Exposure Model	A-  PAGEREF _Toc320006479 \h 
15  

  HYPERLINK \l "_Toc320006480"  2.1.7	Provide Any Other Types of Data to
Assess the Reliability of the Overall Exposure Model or Specific Model
Elements	A-  PAGEREF _Toc320006480 \h  16  

  HYPERLINK \l "_Toc320006481"  2.1.8 	General Comments on Algorithms
and Exposure Assumptions	A-  PAGEREF _Toc320006481 \h  17  

  HYPERLINK \l "_Toc320006482"  APPENDIX B – DETAILED REVIEWER
COMMENTS AND    HYPERLINK \l "_Toc320006483"  EPA RESPONSES:        Part
2 – User Interface 

  HYPERLINK \l "_Toc320006484"  2.2.1 	Provide Comments on the Ease of
Use of the FIAM-PWP Model	B-  PAGEREF _Toc320006484 \h  1  

  HYPERLINK \l "_Toc320006485"  2.2.2 	Comment on the Adequacy of the
Model Help Screens	B-  PAGEREF _Toc320006485 \h  2  

  HYPERLINK \l "_Toc320006486"  2.2.3 	General Comments on User
Interface	B-  PAGEREF _Toc320006486 \h  2  

  HYPERLINK \l "_Toc320006487"  APPENDIX C – DETAILED REVIEWER
COMMENTS AND    HYPERLINK \l "_Toc320006488"  EPA RESPONSES:       Part
3 – Documentation 

  HYPERLINK \l "_Toc320006489"  2.3.1 	Comment on Whether the References
Have Been Accurately Identified	C-  PAGEREF _Toc320006489 \h  1  

  HYPERLINK \l "_Toc320006490"  2.3.2 	Provide Comments on the Model
Documentation (Does It Present the Model Construct, Selected Model
Inputs, and Model Results in a Clear, Complete and Useful Manner?)	C- 
PAGEREF _Toc320006490 \h  2  

  HYPERLINK \l "_Toc320006491"  2.3.3 	Provide Comments on Whether the
Documentation for the FIAM - PWP Make Clear the Sources and Anticipated
Magnitudes of Variability and Uncertainty in Model Results	C-  PAGEREF
_Toc320006491 \h  4  

  HYPERLINK \l "_Toc320006492"  2.3.4 	Comment on Whether the Model
Algorithms are Consistent with the Descriptions in the Model
Documentation	C-  PAGEREF _Toc320006492 \h  5  

  HYPERLINK \l "_Toc320006493"  2.3.5 	General Comments on Documentation
C-  PAGEREF _Toc320006493 \h  6  

 

1 INTRODUCTION

1.1	Background

The U.S. Environmental Protection Agency developed a model for
predicting the potential exposure to formaldehyde associated with use of
pressed wood products in new construction, new homes, and renovations,
the Formaldehyde Indoor Air Model - Pressed Wood Products (FIAM-pwp).
The FIAM-pwp is an adaptation of a model developed during the 1980s by
Dr. Thomas Matthews and colleagues, at the Oak Ridge National
Laboratory. The “Matthews model” was designed to estimate the
steady-state indoor formaldehyde concentration due to emissions from
wood products in a single indoor compartment or zone. Product emission
rates in the model are dependent on the formaldehyde concentration in
the vapor phase – as the concentration increases the emission rates
decrease, other things being equal. The initial (“steady-state”)
indoor concentration calculated by the model is assumed to decrease over
time, as the formaldehyde “reservoirs” in various sources (or sinks)
are gradually depleted. The decrease over time is assumed to follow a
first-order exponential process, at a decay rate that corresponds to an
assumed half life for the collective formaldehyde emissions. In the
mid-1990s, EPA conducted an assessment of the contribution of urea
formaldehyde (UF)-bonded wood products to formaldehyde levels in homes.
Evaluations of then-current EPA exposure models were conducted and
shortly thereafter, a modified version of the Matthews model was
developed. The primary modifications were (1) estimation of steady-state
formaldehyde concentrations for the case of a two-zone indoor
environment, and (2) incorporation of reversible (re-emitting) indoor
sinks. The modified model also included the option to be run in a
single-zone mode. FIAM-pwp is characterized as an upper-tier, empirical
model wherein the model structure is determined by the observed
relationships among experimental data. It is useful to forecast and
describe trends in behavior although it is not necessarily
mechanistically relevant behavior. The dynamic, time-varying behavior of
the state variables in a single run provides deterministic results,
rather than a set of probabilistic outcomes.  

Early in 2009, FIAM was further revised to include additional features
and to make it more user-friendly.  Later in the same year, a peer
review of the model was facilitated for EPA by Eastern Research Group,
Inc. (ERG). The mechanism for the peer review was individual letter peer
reviews conducted following the policy provided in EPA’s Peer Review
Handbook 3rd Edition ( HYPERLINK
"http://www.epa.gov/peerreview/pdfs/peer_review_handbook_2006.pdf"
http://www.epa.gov/peerreview/pdfs/peer_review_handbook_2006.pdf ). The
purpose of the peer review was to review and comment on three main
aspects of the FIAM-PWP model: the soundness of the algorithms and
exposure assumptions; the ease of use of the graphical interface; and
the completeness and clarity of the model documentation. Unedited
written comments submitted by four peer reviewers in response to the
peer review charge were organized in a report prepared for EPA by ERG in
October 2009.

1.2	This Report

This report provides EPA’s responses to the peer review comments,
organized under three main headings:

Part 1 – Algorithms and Exposure Assumptions

Part 2 – User Interface

Part 3 – Documentation

The reviewer comments and EPA’s responses are summarized in Section 2
of this report under these three main headings. Appendices A, B and C
contain detailed extracts of the peer reviewer comments for Parts 1, 2
and 3, respectively, along with EPA’s responses to the comments. 

Since the peer review, the FIAM-pwp model has been revised to produce a
streamlined version (Version 2.0) that runs under EPA’s Internet
Geographical Exposure Modeling System (IGEMS). Based in part on the
reviewer comments, a decision was made to remove certain FIAM features,
including input screens for indoor sinks, an area of considerable
modeling uncertainty, and emission characteristics of cabinet
components, which had significant input demands for the user. In
addition, exposure groups are being handled differently in the revised
version, for consistency with the approach used in EPA’s 2012
formaldehyde exposure assessment.

2 SUMMARY OF PEER REVIEW COMMENTS AND EPA RESPONSES, ORGANIZED BY
CHARGE QUESTION

2.1	Part 1 - Algorithms and Exposure Assumptions

Comments and responses for this part of the peer review are summarized
below, for each of seven charge questions. In addition, general comments
across all questions are provided for one reviewer who chose to provide
comments in this manner rather than organizing them under the specific
charge questions. Detailed extracts of reviewers’ comments and summary
EPA responses related to these charge questions can be found in Appendix
A.

2.1.1 	Comment on the Adequacy of the Mathematical Equations and the
Level of Confidence in Indoor Air Concentrations and Exposure Estimates
Provided by the Model

This particular charge question had the most extensive set of reviewer
comments. Responses to this question account for ten of the 25 appendix
pages listing detailed reviewer comments. Given the extensive comments
to this question, they are grouped into the following six subheadings:

Confidence In Model Predictions

Modeling Of Product Emissions

Modeling of Emissions Decay

Modeling of Indoor Sinks

Modeling of Exposure, Including Product-Specific Exposure

Miscellaneous Comments

	

Confidence in Model Predictions. The reviewers generally expressed some
confidence in the model predictions but also had some reservations, as
evidenced by the following verbatim comments:

Reviewer 1 – Based on the data provided and the assessment analysis,
the model should provide reasonable conservative estimate of the
formaldehyde exposure given the relatively long half-life time.

Reviewer 2 – Regarding the level of confidence one might have in the
model’s prediction of actual indoor concentration and exposure levels,
my sense is that they should be considered accurate within a factor of
2.

Reviewer 3 – The mathematical/algorithmic formulation and the
computational implementation of the model under review are adequately
explained, with sufficient clarity and level of detail; the
computational implementation of the model also appears robust and
accurate. Nevertheless, given the simplifying assumptions in the
formulation of the model, and the significant variabilities inherent in
the data sets that were used to derive parameter values for FIAM-pwp,
the level of confidence in the indoor air concentration and exposure
estimates provided by the model, should in general be considered, at
most, medium.

Reviewer 4 – In summary, I concur with the calculation of initial
concentration (high confidence in result with reliable chamber test
data; however, the treatment of sinks and average daily concentration
for chronic exposure appear to require some revisions to improve the
quality of the result (low confidence in result).

Reviewer 5 -- The analysis provided in Section 5.3 of the documentation
suggests that whatever the convolutions involved, the model gives
respectable predictions. My reading of the tables in that section, which
summarize modeling results compared to measured values in the EPA pilot
study house, is that the model ranged from about 24% low to about 65%
high. Predictions within that range are quite good, given the
uncertainties in input data and unavoidable modeling assumptions.

Modeling of Product Emissions. Three reviewers commented in this area.
The first reviewer noted that Equation 3-5, which forms the basis for
the emission rate modeling engine, appears to be a reasonable
approximation of reality and is in complete agreement with his previous
work on “backpressure” modeling. He further suggested that the
analytical measurement of Csat – the saturation concentration of
formaldehyde in the air over the emitter of interest – could provide a
potentially valuable extended methodology for the evaluation and
development of the model’s parameters relative to specific emitters.
EPA agrees that a well designed analytical study could be useful to
provide information on the saturated air concentration of formaldehyde
and has noted it in the model’s documentation as a suggestion for
interested researchers.  However, such a study is not within the scope
of this project. The new version of the model includes the calculated
equilibrium concentration as an intermediate output that is displayed on
the Source Screen for each source selected or entered by the user.

The second reviewer expressed the following opinion:

There appears to be some confusion between diffusion and convective mass
transfer. The model is based on convective mass transfer over the
surface, which is not based on the Fick’s Law of diffusion. However,
it is more likely that, except for the initial emission period, the
emission rate will be controlled by the internal diffusion process as
suggested by many studies. So the “slope” values derived from the
measured data may have included the combined effects of convective mass
transfer and internal diffusion resistances, resulting in reasonable
predictions. At the “steady state” (about 30 days from initial
loading), the internal diffusion resistance should be the dominant
resistance for formaldehyde emissions even for particleboard. Discussion
of diffusion and convective mass transfer, along the lines of the
reviewer’s comments, will be added to Section 3 of the model
documentation, along with related comments from another reviewer who
referred to a conceptual model published by Mølhave et al. (1995).
According to this model, three conceptual exponential decay compartments
can be attributed to three potential sources of formaldehyde emissions:

Free formaldehyde (short-term compartment);

Decay of paraformaldehyde and other complex structures (medium-term
compartment); and

Hydrolysis of the resin (long-term compartment).

The third reviewer made several observations and suggestions:

The original Versar/EPA model (referring to a version developed during
the 1980s) presented the linear regression parameters for both the
Matthews model and the Hoetjer model (identified as the HBF model in
this early version), as well as the steady-state equations for both
models. It can be shown these are identical models – the difference in
models is in the regression equation used to derive the parameters, not
the theoretical construct. Because slightly different results are
obtained depending on the regression model used and the Hoetjer model is
widely cited in the literature, it is recommended that the model
documentation be revised to include discussion of both approaches. 

If only the Matthews model is retained, then an example calculation
should be provided in the model documentation showing how the user can
start with the HBF model parameters and convert these to the equivalent
Matthews model. Another option would be to provide the HBF equilibrium
concentration as supplemental output, to enable an intuitive
determination about whether a particular product is likely to act as a
sink or a source.

Some pressed wood formaldehyde “emitters” could theoretically act as
sinks under certain conditions based on the properties of wood and the
Matthews model (e.g., products with a very low equilibrium concentration
or, equivalently, low Matthews model intercept). Therefore, there may be
some usefulness in presenting the emission rates by product in the
results, with a negative emission rate indicating a net sink effect.

Adding the Hoetjer model is beyond the scope of the current EPA effort
but discussion of the two approaches, as well as the example calculation
suggested, has been  included in the model documentation. In addition,
as noted above, the equilibrium concentration for each emitter has been
added as an intermediate model output; this output is intended to assist
model users in identifying, for example, a relatively weak emitter that
may act as a sink or “net absorber” in the presence of other,
stronger emitters. An example has been added to the model documentation
to demonstrate how the user can identify a source that is acting as a
sink in the presence of other sources, by running the model with and
without the weak source (if a “source” is acting as a sink, then the
modeled concentration when that source is included along with the
stronger emitters will be lower than when the source is excluded). A new
model feature, enabling users to check or uncheck sources for inclusion
in or exclusion from a model run, makes this procedure very
straightforward to apply. 

Modeling of Emissions Decay. Three reviewers commented on this area. The
first reviewer noted that the approach of exponentially decaying the
indoor concentration is unconventional, but practical considering the
many uncertainties involved in real-life situations, most notably
choosing the right half-life. He supported the idea of relying on
measured field data to estimate the decay of formaldehyde concentration
and to be conservative in the estimation, as he believes the model to
be. The second reviewer expressed the opinion that the value of the
information that has been incorporated in FIAM-pwp would be greatly
enhanced if its formulation explicitly considered decaying emission
rates and provided appropriate formulas for implementing them (either
within FIAM-pwp or in conjunction with other, more comprehensive, indoor
air quality and inhalation exposure systems). The third reviewer
suggested that basic research on the decay profile of individual
products should be encouraged in the model documentation (rather than
exclusively relying on the historical data from the 1980s). The reviewer
suggested that, at a minimum, the model should be revised to allow
consideration of periods of rapid and slow decay in the emission rate as
observed by several researchers.

In response to the comment, discussion has been added to the model
documentation regarding how rapid and slow emissions decay could be
handled via a series of related model runs. The suggestion regarding
basic research on the decay profile of individual products has been
incorporated in Section 3 of the revised model documentation. At this
time, EPA is not revising FIAM-pwp to decay emission rates rather than
concentrations, or to explicitly allow periods of rapid and slow
emissions decay within a single model run, because (1) the underlying
mathematics for decaying emission rates rather than concentrations are
relatively complex, and (2) data to support user choices for emission
parameters governing periods of rapid and slow decay, as well as the
point in time when a transition between the two periods occurs, are
relatively sparse. In addition, the model as designed achieves the goal
of supporting EPA exposure assessments.

Modeling of Indoor Sinks. One reviewer commented extensively on indoor
sinks, expressing some concerns with how sinks are handled in the model,
including the following specific comments:

The treatment of indoor sinks in the current EPA model is problematic
because absorption and desorption of formaldehyde from indoor sinks is a
transient process over relatively short time scales not well suited to a
steady-state or pseudo-steady-state model construct.

It appears that the guidance for the sink intercept given on EPA model
guidance page 3-7 was developed by setting intercept b = ka x Cindoor
assuming dM/dt = 0. When there is a linear relationship between
accumulated mass and indoor air concentration at equilibrium conditions,
dM/dt must equal zero. Therefore, deviations of the intercept from the
calculated value would imply a sustained loss of equilibrium, violating
the assumption used in assigning the sink slope and intercept in the
Matthews equation. Furthermore, the guidance provided for selection of
the equilibrium intercept for indoor sinks appears to treat this
parameter as calibration or adjustment factor rather than as a
predictive input.

To correctly account for net adsorption on these short time scales
within the framework of a pseudo-steady-state model, a non-zero value of
dM/dt must be calculated using Equation 3-13 and set equal to the
Matthews emission rate equation to derive the intercept. This value of
dM/dt will vary as a function of time. Such an approach would
appreciably increase the level of complexity needed to assign the sink
term because indoor concentration and mass storage vary as a function of
time. If an estimate of the indoor concentration profile is available as
a function of time, it should be possible to determine M(t) and dM(t)/dt
as a function of time by numerical integration or a finite difference
model. However, at 30 days, dM/dt may be sufficiently small that a
reasonable default recommendation in the model would be to neglect the
effect of sinks.

EPA has carefully reviewed the reviewer comments relating to sinks, such
as those provided above, and based on its review has removed the
indoor-sinks module from the revised FIAM-pwp model. The supporting data
needed to establish a mathematical model of sink effects for each
material, let alone a combination of materials, are not readily
available and development of such data or conduct of related research is
beyond the scope of this project. Users of the model can incorporate
sinks within the indoor-source module, recognizing the limitations of
that approach. (The mathematical treatment and input requirements for
sinks, when they were included in the model, were identical to that used
for sources.)  As the science regarding sinks advances, EPA could
consider adding sinks to the model in the future.

Modeling of Exposure, Including Product-Specific Exposure. Two reviewers
provided several comments in this area. The first reviewer noted
situations for which the model either is inconsistent or provides
erroneous results:

Relative to the estimation of exposure there appears to be logical error
in the defaults. When one invokes the exposure defaults, the 5 age
groups for children in the model all show 10 years for the exposure
period. Of course, this is impossible for any of the age groups for
children which are all less than 10 years in length. This is explained
in the documentation and in the online help button; however, I found it
somewhat confusing. My suggestion is that the default shown in the
program for any age group should be the maximum time a person can spend
in that age group.

The program allowed me to put in a single day for the duration of the
exposure but the results indicate that a calculation error occurs for
this input. The calculated ADC is above the initial concentration and
the time percentages come out to be >100%. If this error is not easy or
practical to fix then explanation should be included advising the
minimum time for duration of exposure.

These potential concerns were addressed by handling exposure groups and
exposure durations differently in the revised FIAM-pwp. The model now
has six default exposure groups covering different age ranges, and the
model results include year-by-year ADC calculations for each exposure
group, for the first eleven years of the model run, such that the user
can select the ADC for the time frame of interest.

The second reviewer commented extensively on exposure estimation in
general and product-specific exposure in particular:

The usefulness of the model could be strengthened by discussing the
incremental contribution of added products because regulations, if any,
are likely to be developed on a product-specific basis. The model
documentation and results emphasize the aggregate contribution of all
products (and background sources), and running the model with and
without a product of interest can produce counterintuitive results. The
model documentation should focus on both the determination of
incremental exposure attributable to a specific product and total
exposure for a loaded home. 

Assessment of acute/chronic exposure is associated with different
uncertainties and conceptual models. In the acute setting, the
back-pressure effect may attenuate the contribution of a new product
such that the total acute indoor exposure concentration is not affected.
However, in a chronic or long- term timeframe, it is reasonable to
expect that the majority of available formaldehyde and precursors
ultimately will partition from the product into the indoor space. My
recommendation is that a conceptual framework be developed/documented to
ensure that users take these factors into consideration when
investigating acute versus chronic timeframes. For the acute timeframe,
it is important to consider the total collection of assembled indoor
products. For chronic exposure, a more accurate characterization of dose
attributable to a single product is likely obtained by considering the
decay profile of that product in the absence of other products; the
attenuation due to the backpressure effect is temporary, and eventually,
a source that is initially attenuated will begin to emit formaldehyde as
the indoor-air concentration decreases. 

For acute exposure, an option could be provided to identify ‘baseline
products’ and ‘added products’.  Model outputs of exposure and
product-specific emission rates would be presented with and without the
added products. I believe this would greatly assist users of the model
in intuitively understanding the backpressure effect, as well as assist
specific manufacturers, regulators and other interested parties in
understanding the incremental effect of specific products on indoor air
levels.

The incremental chronic exposure attributable to a specific product or
collection thereof should be assessed, rather than the aggregate
cumulative exposure for all indoor products. Because the typical
situation will reflect products of various ages, the chronic dose
attributable to a specific product seems to be a more relevant metric
than the total indoor concentration over time assuming an initial
loading.

Under the current model guidance, the true increased cancer risk
attributable to installation of a specific product would not be reliably
determined. Initially, the backpressure effect will suppress a portion
of formaldehyde emissions, such that if the user were to compare
cumulative dose with and without the product the lifetime dose for that
product will be underestimated. If the equilibrium concentration (i.e.,
the theoretical concentration in the absence of ventilation) of a
product is less than the pre-existing steady-state concentration, then
running the model with and without the product would result in the
physically impossible conclusion of negative incremental cancer risk
because the product would initially act as a net sink.

EPA has modified the model to address this comment; an example has been
added to the model documentation to demonstrate how the user can
identify a source that is acting as a sink in the presence of other
sources, by running the model with and without the weak emitter. The
model now has considerable flexibility in the source inputs, allowing
the users to check or un-check sources such that individual products or
groups thereof can be examined from both short- and long-term
perspectives.

Miscellaneous Comments. Reviewers had several additional comments or
suggestions that did not fall under any of the above categories:

It is suggested that the modelers consider a modeling program based on
The Advanced Continuous Simulation Language (ACSL) computer language.
ACSL easily handles simultaneous differential equations while providing
parameter estimates from data via optimization and an intuitive
graphical interface.

The uncertainty and variability levels in the available data strongly
suggest the need for a distributional/probabilistic model formulation
that explicitly employs available data to develop ranges rather than
point estimates of concentrations and exposures.

One concern is the interaction between ventilation rate and decay rate.
Given long-term testing data, the total reservoir of formaldehyde and
formaldehyde precursors for a given ventilation condition could be
calculated. Therefore, in the future when decay characteristics are
better understood, it should be possible to adjust for ventilation rate
and account for total mass of formaldehyde generated over the lifetime
of the product.

The model should provide sufficient ancillary information to understand
the sensitive parameters and identify other potential areas for testing.


With regard to the programming language, Version 2.0 of the FIAM-pwp has
changed substantially from the peer-reviewed version and has been
incorporated into an existing web-based platform, IGEMS. The new
web-based version has a much quicker response time, particularly when
users are providing inputs. The sensitivity/uncertainty discussion in
the FIAM-pwp v2.0 documentation has been bolstered, including an
appendix containing the sensitivity analysis that was performed as part
of EPA’s 2012 formaldehyde exposure assessment for illustration
purposes. EPA acknowledges that Monte Carlo simulation, or a similar
probabilistic tool, would be useful given the many combinations and
permutations that an assessment of formaldehyde from pressed wood
products can entail. However, such an enhancement is not within the
current scope of this project.

2.1.2 	Comment on the Default Scenarios and Assumptions Used by the
Model

The following specific suggestions were provided by reviewers in
response to this charge question:

The modelers should consider using and including the seminal study by
Murray and Burmaster (1995) as an important reference for potential use
with this model. 

First, it should be made clear that many of the default emissions
parameters in the model are based on historical testing and may or may
not be representative of current products. Second, clear guidance should
be provided to end users (e.g. importers, manufacturers, non-profit
organizations or regulators) about the data and steps required to assess
the HCHO incremental exposure attributable to the introduction of a
specific product into the home.

EPA agrees with these comments and has revised the FIAM-pwp
documentation accordingly. The Murray and Burmaster article, referenced
as one of the sources considered in choosing an air exchange rate for
the recent formaldehyde exposure assessment, has been cited in the model
documentation. Some discussion relating to historical testing, as well
as an example relating to incremental exposure, has been added to the
model documentation, keeping in mind that an assessment of incremental
exposure is not a primary intent of the model. Additional reviewer
comments regarding the historical data underlying default emissions
parameters are included under the next charge question, relating to data
sources for the model. 

2.1.3 	Comment on the Data Sources

In response to this charge question, one reviewer reiterated his comment
regarding the utility of the Murray and Burmaster data in choosing a
residential air exchange rate. Two of the reviewers commented that the
data underlying emissions parameters are dated and, thus, not
necessarily reflective of current or recent manufacturing practices. One
reviewer commented on the airflow rates and underlying data for the
default house selections labeled as Conventional 1 and Conventional 2.
These houses are no longer included as default selections in the model;
they have been replaced by five generic structure types –
single-family attached and detached homes, apartments,
mobile/manufactured homes, and camper trailers – as detailed in the
revised FIAM-pwp (v2.0) User’s Manual and Documentation report. The
same reviewer commented that use of the Langmuir isotherm for wallboard
as a sink should be further explained or justified. The Agency has
modified the model; sinks are no longer explicitly included due to a
number of uncertainties, such as those noted in the peer reviewer
comments under the first charge question above (see Section 2.1.1). 

2.1.4 	Comment on Whether the Data Sources Provide Unused Information
That Would Assist in Quantifying Uncertainty or Variability for
Subsequent Model Use

The reviewers had several suggestions in response to this charge
question:

Although the data sources are incomplete for quantifying the
uncertainty, it is possible to conduct an analysis using error
propagation theory with assumptions on the uncertainties in the raw
data.

The Murray/Burmaster database on air exchange rates in U.S. residences
could assist in gauging the variability of this parameter. Also, I
believe the current data used in the model on emission sources discloses
a level of uncertainty and variability that could also be mined by a
person who is knowledgeable about these data and their origins.

The variability in the available data could be used to derive
distributional estimates of emission rates that would support
probabilistic (and population based) indoor concentration and
corresponding exposure estimates with explicit characterization of
uncertainties and variabilities.

The uncertainty and sensitivity analysis presented in Matthews et al.
1985 could be discussed.

EPA acknowledges that probabilistic tools would be useful to address the
many possible combinations/permutations of inputs that an assessment of
formaldehyde from pressed wood products could entail. However, as noted
previously, such an enhancement is not within the current scope of this
project. The uncertainty and sensitivity analysis presented by Matthews
et al. has been summarized as part of an appendix on model sensitivity
that has been added to the revised model documentation.

2.1.5 	Comment on Whether the Model Input Variables and the Estimates
Assigned to Them Appropriately Reflect Variability

The reviewers had several suggestions for addressing variability,
ranging from use of high- and low-end model inputs (e.g., emissions
strength) to a distributional/probabilistic approach. The Agency has
incorporated additional text addressing these comments in the model
documentation. For example, the model sensitivity analysis that was
described in EPA’s 2012 formaldehyde exposure assessment report has
been included in an appendix to the revised model documentation. Other
reviewer comments or suggestions related to uncertainty or variability
are as follows:

The single exponential equation for decay is consistent with the
historical cross-sectional residential data but inconsistent with
studies that have investigated single products or collection of products
over time. The true area under the curve for dose given a long enough
exposure time is likely better represented by the two- or three-
compartment exponential models than the single exponential model because
these models better represent the decay in the reservoir of formaldehyde
precursors in a specific product or collection of products.

EPA should perform a sensitivity analysis to determine whether
measurement of the mass transfer coefficient is a data need or data gap
when newly acquired chamber test data are entered.

The documentation does not provide a credible method for selecting the
indoor sink parameters.

The model screens and documentation should make it clear that
Conventional 1 represents a single unoccupied test home. The analysis
used to derive Conventional 2 should be described.

The degree to which the historical data in Table 2-2 reflects domestic
and imported production is not known.

EPA agrees that the first two suggestions, regarding formaldehyde decay
and the mass transfer coefficient, are of interest; however, conducting
additional research and analysis is beyond the scope of the current
effort, and the model as designed meets the Agency’s needs for
exposure assessment. The next two comments/suggestions have been
addressed by (1) removing indoor sinks in the revised model, due to a
number of uncertainties that may lead to misleading results if sinks
were included in model runs (see Section 2.1.1), and (2) replacing the
houses denoted as Conventional 1 and Conventional 2 with five generic
structure types (see Section 2.1.3). EPA has addressed the final comment
by noting this uncertainty in the model documentation.

2.1.6 	Comment on the Adequacy of the Documentation Supporting the
Overall Plausibility of the Exposure Estimates Generated by the Exposure
Model

Several reviewers suggested that model estimates could be compared with
measurement results from prior field studies to better evaluate their
plausibility. Some other reviewer comments or suggestions related to the
plausibility of exposure estimates are as follows:

Comparison of model estimates with relevant observational data sets for
indoor-outdoor relationships and of personal level (e.g. from studies
such as RIOPA – see, e.g., the Liu et al., 2006, reference in the
Addendum of this review) would be valuable in assessing, at least
qualitatively, the plausibility and relevance of the exposure estimates
derived from FIAM-pwp.

EPA has not shown how use of an overall residential half-life is to be
used to derive product specific dose. Running the model with and without
the product of interest will not result in meaningful conclusions as
discussed previously. Accordingly, there is great uncertainty in product
specific doses under the current model framework.

EPA has not considered whether the historical 2.92-year cross-sectional
half-life is representative of the decay profile of modern products or
the potential variability in the formaldehyde reservoirs and decay
between products.

EPA has addressed the comments on half life by the inclusion of
additional text in the model’s documentation. With regard to the last
comment, EPA has considered the possibility that a value of 2.92 years
for the half life may be too high; the default value for the half life
in the revised version (v2.0) of FIAM-pwp is 1.5 years. A section has
been added to the revised documentation on using the model to estimate
the average concentration in a recent California field monitoring study
that focused on newly constructed residences, for comparison with the
concentration that that was measured in the study.

2.1.7 	Provide Any Other Types of Data to Assess the Reliability of the
Overall Exposure Model or Specific Model Elements

One reviewer reiterated an earlier comment (see Section 2.1.1) regarding
the potential utility of Csat as a critical emitter variable. A second
reviewer reiterated an earlier comment concerning a framework for
assessing product-specific, long-term exposure (again see Section
2.1.1). EPA has added text to the model documentation in response to
these comments. A third reviewer noted that there are various studies
and related data sets, both from the U.S. and abroad, that have appeared
in recent years and could potentially be used for evaluation of specific
FIAM-pwp elements; one such data set has been used for that purpose, as
noted above. The reviewer also provided an addendum containing a number
of literature references pertaining to such studies and data sets. The
references have been evaluated and the applicable ones have been
incorporated in the revised model documentation.

2.1.8 	General Comments on Algorithms and Exposure Assumptions

One reviewer provided general comments rather than replying to the
specific areas above:

Seems like a convoluted way to model emissions: The dependent variable
of primary interest – the HCHO concentration – is inherent in the in
the prediction algorithm; i.e., it is also used as an independent
variable. Why not just use the intercept values and simplify the model?

That said, the analysis provided in section 5.3 of the documentation
suggests that whatever the convolutions involved, the model gives
respectable predictions.

Regarding the selection of a decay constant (or half-time) for
predicting the rate at which the HCHO concentration will decrease, I
suspect the default choice of k = 0.23 yr-1 (or t1/2 = 3 years) might be
improved by having different values for different time periods.

One significant shortcoming of this model, and many indoor models that
assume complete mixing within compartments, is that it doesn’t deal
with “intimate” sources such as toys, cribs, bedding materials, and
clothing. I think the most straightforward approach to such sources is
to assume a smaller compartment or zone around such sources for the
purpose of predicting concentrations.

Other reviewers had no difficulty with the dependent variable being
inherent in the prediction algorithm. EPA agrees that allowing for
different decay values for different time periods is a reasonable
suggestion; however, the modeling framework needed to accomplish this is
not trivial and would require a resource-intensive analysis to supply
appropriate values. As noted in Section 2.1.1 (under the heading
Modeling of Emissions Decay), discussion has been added to the model
documentation regarding how rapid and slow emissions decay could be
handled via a series of related model runs. The FIAM-pwp enables the
user to model intimate sources using a two-zone modeling approach
whereby one of the zones is used to represent the “virtual air
space” surrounding an individual exposed to an intimate source.

2.2	Part 2 - User Interface

Comments and responses for this part of the peer review are summarized
below, for each of two charge questions along with general comments
across all questions for one reviewer. Detailed comments and responses
related to these charge questions can be found in Appendix B.

2.2.1 	Provide Comments on the Ease of Use of the FIAM-PWP Model

The reviewers generally felt that the user interface was relatively
clean, straightforward and easy to use. Several reviewers also commented
that the online version is very slow whereas the local version, although
faster, requires adjusting local computer settings in ways that may be
challenging for some systems and users. These latter comments, as well
as some comments regarding specific aspects of the interface, were
addressed by developing an online version of FIAM-pwp that is run under
IGEMS with a redesigned, faster responding user interface. Some specific
reviewer comments/suggestions that have been addressed in the revised
user interface and/or model documentation are as follows:

It would be useful to provide some sort of guide for users to step
through the modeling process (such as step 1, 2, 3, ….)

Maybe some “warnings” regarding the level of uncertainties in the
calculations could be incorporated, as reminders, in the data-entry
screens.

Replacing this field (referring to air change rate) with plain text
would make clear that this value cannot be edited.

The codes in the model do not match Table 2-2 in all instances.

Consider writing a routine to allow the information to be printed
directly from the web-browser (referring to Excel output that is an
option with the local version of the current model).

2.2.2 	Comment on the Adequacy of the Model Help Screens

The reviewers generally felt that the help screens are intuitive and
adequate in content. There were some comments related to facets of the
“audio help” that did not appear to work, but this model feature
will not be included in the version (v2.0) that will be running under
IGEMS. One specific reviewer suggestion regarding help screens that has
been addressed is as follows:

The help screens could be enhanced by including some statements
regarding the level of precision expected for some of the entries.

2.2.3 	General Comments on User Interface

One reviewer provided general comments rather than replying to the
specific areas above:

More careful attention to significant figures is needed.

On the Sinks Screen, it seems counter-intuitive that the B value is
positive rather than negative.

The first comment has been addressed in the revised version of the
model. The second comment is no longer applicable because sinks are
being removed in the revised version, due to a number of uncertainties
that may lead to misleading results if sinks were included in model runs
(see Section 2.1.1).

2.3	Part 3 – Documentation

Comments and responses for this part of the peer review are summarized
below, for each of four charge questions along with general comments
across all questions for one reviewer. Detailed comments and responses
related to these charge questions can be found in Appendix C.

2.3.1 	Comment on Whether the References Have Been Accurately Identified

The reviewers had several suggestions for additional references; in one
case a reviewer provided a long list of possible references. The
following are specific reviewer suggestions for references:

The fourth reference (the CARB Final Regulation Order…) should be
identified as available at:   HYPERLINK
"http://www.arb.ca.gov/regact/2007/compwood07/fro-final.pdf" 
http://www.arb.ca.gov/regact/2007/compwood07/fro-final.pdf . The 8th
reference (D. Hare…) can be found at, and should be identified as
available at:   HYPERLINK
"http://www.ecobind.com/research/Evaluating_the_Contribution.pdf" 
http://www.ecobind.com/research/Evaluating_the_Contribution.pdf .

Any and all the references, particularly from the EPA, GEOMET, Versar or
industry groups, should be identified by their web links, if available,
and if not available on the web, should be considered for posting online
with a link.

The two additional references (Murray and Burmaster on residential
ventilation rate database and Jayjock on “Backpressure” modeling)
should be considered for inclusion.  

For more information regarding formaldehyde emissions from pressed wood
products, model users could be referred to a U.S. EPA (1996) ORD report
– Sources and Factors Affecting Indoor Emissions from Engineered Wood
Products.

It may be helpful to provide citations to studies that have summarized
recent measured formaldehyde indoor air concentrations in new and
existing structures. Examples include Hodgson and Levin 2003 and Hodgson
et al. 2000.

Groah’s white papers and reports have been invaluable in my past work
on formaldehyde. However, I recommend that these references be replaced
with peer-reviewed publications, conference proceedings, or governmental
citations. Examples of publications addressing the long-term decay of
formaldehyde emissions from consumer products include: Mølhave et al.
1995; Brown 1999; Zinn et al. 1990; Myer and Hermanns 1985; and CARB
2007.

EPA has made all references used for development of this model available
in the model documentation. Additionally, some of the suggested
references have been added to the revised model documentation.

2.3.2 	Provide Comments on the Model Documentation (Does It Present the
Model Construct, Selected Model Inputs, and Model Results in a Clear,
Complete and Useful Manner?)

The reviewers had a variety of specific suggestions/comments relating to
model documentation:

The model is not truly a steady-state one since the emission source
decays over time. It is a “quasi-steady state” model in which the
concentration at a given calculation point was treated as “constant”
for the purpose of simplifying the calculation without losing too much
accuracy.

Most data provided on formaldehyde emissions are from early studies.

Validity of the exponential decay model is questionable. Recent data
have shown that the power-law model would be more suitable; secondary
emissions due to hydrolysis are also possible.

The Fick’s law is for diffusion, while here the process is governed by
convective mass transfer across the boundary layer.

The definition of vapor phase concentration and bulk phase concentration
is unclear. I see no need to define a vapor phase concentration since,
by default, we are always dealing with the vapor concentration in the
air or in the material in this model.

For me it would be more logical to change the order of Chapters 2 and 3.

The use of “D” on page 3-4 to be unnecessarily complex.

The main issue with the documentation is the justification of the
formulation of certain elements of the model per se (e.g., the
assumption of constant rather than decreasing emission rates, etc.)
rather than of potential alternatives.

The treatment of sinks is not transparent and confusing because the rate
of storage changes with time but the model used is a steady-state model.
The approach used appears to be more an attempt to modify model results
to reflect the measured concentration than a rigorous handling of sinks.

Some stakeholders may not be familiar with the linear regression
protocol required to estimate the parameters for the Matthews and
Hoetjer formaldehyde mass transfer models. The documentation could be
revised to show an example linear regression for the data from Koontz et
al. (1996).

The documentation should explain the relationship between the Matthews
and Hoetjer equation. The Hoetjer equation or equivalent mass transfer
equation is frequently cited in the literature and some stakeholders may
not be familiar with the Matthews equation.

Consider including in an appendix the analysis of the EPA Pilot Study
data used to determine the mass transfer coefficient and intercept (p.
5-3).

EPA has addressed most of these comments in Section 3.0 of the revised
model documentation. As noted previously, indoor sinks have been removed
in the revised version, due to uncertainties such as the one noted above
that may lead to misleading results if sinks were included in model
runs. Discussion in this regard has been included in Section 3.0 of the
revised documentation.

2.3.3 	Provide Comments on Whether the Documentation for the FIAM - PWP
Makes Clear the Sources and Anticipated Magnitudes of Variability and
Uncertainty in Model Results

Several reviewers indicated that sources of uncertainty were addressed
well in the model documentation. One reviewer felt that a table
explicitly listing various sources of uncertainty and variability
associated with specific processes and parameters of the model, along
with a factor or range specifying the approximate magnitude of each
uncertainty/variability, would be extremely useful. Ideally, the
reviewer would like to see a probabilistic “re-formulation” of the
model, with explicit discussion in the documentation of the appropriate
distributions of values for the various inputs. Another reviewer
suggested a simple Monte Carlo analysis as an appendix to provide some
guidance on whether refinement of a particular input parameter is likely
to change the conclusion. EPA agrees that more information on how the
each input parameter can affect the results would be useful; for this
reason, the sensitivity analysis that is part of EPA’s 2012
formaldehyde exposure assessment report has been added to the model
documentation. The same reviewer also suggested incorporating a mass
transfer coefficient sensitivity analysis into the model or model
documentation, along with a discussion of the uncertainty and
sensitivity analysis presented in Matthews et al. 1985. As noted
previously in this document (see Section 2.1.4), a summary of the
Matthews analysis also has been added to the model documentation.

2.3.4 	Comment on Whether the Model Algorithms are Consistent with the
Descriptions in the Model Documentation

Reviewers generally agreed that the model algorithms are consistent with
their descriptions in the model documentation. One reviewer verified the
accuracy of model calculations by comparing results with those obtained
from a different indoor-air model with the ability to handle
backpressure effects. The same reviewer noted that the slope in the
Matthews equation is not consistently defined, referring to pages 2-11
and 3-4 of the model documentation, and he also noted that the model
documentation is not clear regarding whether the background
concentration is included in the ADC and LADC calculations. Another
reviewer stated a preference for R2 values rather than the correlation
coefficients included in Table 2-2. EPA has added clarifying text to the
documentation to address the points made in the referenced pages and
table.  

2.3.5 	General Comments on Documentation

One reviewer provided general comments rather than replying to the
specific areas above:

I found the documentation to be generally clear and quite useful. 

The slope should not be described as “...in the form of a first-order
sink representation.”

The phrase “...flows of formaldehyde mass into and out of a fixed
volume must be equal...” is not correct. In fact, equation 3.1
contradicts that.

While the term “half-life” has a certain value because of its
understandability, the normal way to determine it is from the decay
constant.

One area that could be significantly improved is the mathematical
introduction to the model.

EPA has revised the above-referenced text to include more clarity and
transparency in the revised model/documentation, except for the
statement relating to the first-order sink representation, which is no
longer applicable because sinks have been removed from the model.



APPENDIX A – DETAILED REVIEWER COMMENTS AND

EPA RESPONSES: Part 1 – Algorithms and Exposure Assumptions

2.1.1 Comment on the Adequacy of the Mathematical Equations and the
Level of Confidence in Indoor Air Concentrations and Exposure Estimates
Provided by the Model

Reviewer / Comment	Response

Zhang – The mathematical equations are adequately explained except
that there appears to be some confusion between diffusion and convective
mass transfer. The model is based on convective mass transfer over the
surface, which is not based on the Fick’s Law of diffusion. The air
velocity over the surface would have significant effect on the “mass
transfer coefficient”, and hence the model parameter, “slope”. The
values of the “slope” appear to be small (maximum of 3.0 in the case
of painted dry wall). This is likely that the most data are from small
scale chamber tests where air velocity over the material surface is
small compared to the typical airflow conditions in ventilated rooms.
Therefore, the model may underestimate the emission rates if the
emission is indeed controlled by the convective mass transfer process.  
EPA has added clarifying text to  the revised documentation and/or help
screen text

Zhang – However, it is more likely that except for the initial
emission period, the emission rate will be controlled by the internal
diffusion process as suggested by many studies including those cited in
the report for pressed wood products. So the “slope” values derived
from the measured data may have included the combined effects of
convective mass transfer and internal diffusion resistances, resulting
reasonable predictions. At the “steady state” (about 30 days from
initial chamber test), the internal diffusion resistance should be the
dominant resistance for formaldehyde emissions even for particleboard.
To understand and proof this point, a more detailed numerical simulation
can be made for homogeneous sources as assumed in the study (i.e., there
were free formaldehyde uniformly distributed in the material).	EPA has
added clarifying text to the revised documentation and/or help screen
text

Zhang – After establishing the steady state initial concentration
based on Mathew’s model, the concentration in a zone that has a higher
concentration is assumed to decay exponentially over time, given a half
life time as the input parameter. This approach is unconventional, but
practical considering the many uncertainties involved in real situation.
The challenge would be to choose the right the half-life time.	No
response/action required

Zhang – Formaldehyde may not be distributed uniformly inside the
material. They may be more concentrated in resin spots and less in the
wood particles/chips. In addition, hydrolysis is possible for additional
formaldehyde generation within the material, resulting longer and higher
emission rates in the long term. This is still a research topic and no
reliable model has been developed yet. Therefore, I support the idea of
relying on measured field data to estimate the decay of formaldehyde
concentration, and be conservative in the estimation, as intended by the
model.	No response/action required

Zhang – Based on the data provided and the assessment analysis, the
model should provide reasonable conservative estimate of the
formaldehyde exposure given the relatively long half-life time.	No
response/action required



 

Reviewer / Comment	Response

Jayjock – Regarding the appropriateness of the algorithms and
assumptions, Matthew’s basic approach appears to be based on solid
first principles of conservation of mass and the validity of diffusion
via Fick’s Law. Equation 3-2 appears to be relatively simple,
self-evident and quite appropriate.	No response/action required

Jayjock – Similarly, equation 3-5 forms the basis for the emission
rate modeling engine. It is well explained and appears to me to be a
reasonable approximation of reality. This equation is in complete
agreement with previous work that I have done on “backpressure”
modeling.	No response/action required

Jayjock – Equation 3-5 is essentially identical to an algorithm that I
have often used for backpressure modeling (shown to the right):

Csat = the saturation concentration (mg/m3) of formaldehyde in the air
over the emitter of interest.  That is the concentration that would be
expected in a volume (with high surface area/volume ratio) in which the
emitting walls were made up entirely of the emitter of interest and
there is zero ventilation.

Note the equivalence of this last equation to 3-5 above

and the identity of m (shown t the right):

Clearly, the relatively straightforward analytical measurement of Csat
could provide a potentially valuable extended methodology for the
evaluation and development of the model’s parameters relative to
specific emitters.	EPA agrees that Csat is a useful interpretation aid,
but measurements such as those suggested here are outside of the scope
of this project; the suggestion as been noted in Section 3 of the model
documentation 

Jayjock – I believe that the treatment of indoor sinks as developed by
and explained in the model represent a real step forward in indoor
modeling science. I would suggest that the author of the documentation
in this area expand the explanation to include some example data and
treatment that illustrates the estimation of parameters for the Langmuir
model. That is, the last paragraph on page 3-7 supplies the estimated
values but my sense is that the explanation would be significantly more
lucid if the actual procedure and values used in the fitting exercise
were demonstrated and included.	After consideration of all peer review
comments received, indoor sinks have been  removed from the model
because of  modeling uncertainties (see Section 2.1.1)

Jayjock – Regarding the level of confidence one might have in the
model’s prediction of actual indoor concentration and exposure levels,
my sense is that they should be considered accurate within a factor of
2. 	No response/action required

Jayjock – Relative to the estimation of exposure there appears to be
logical error in the defaults. When one invokes the Exposure Defaults
within the model, the 5 age groups for children in the model all show 10
years for the exposure period. Of course, this is impossible for any of
the age groups for children which are all less than 10 years in length. 
This is explained in the documentation and in the online help button;
however, I found it somewhat confusing. My suggestion is that the
default shown in the program for any age group should be the maximum
time a person can spend in that age group.	The model now has six default
exposure groups covering different age ranges, and the model results
include year-by-year ADC calculations for each group

Jayjock – It is not difficult to imagine relatively high doses of CH2O
potentially causing biological damage over a period of a few days. As
such, the estimation exposure in the Results Screen of Average Daily
Concentration (ADC) and Average Daily Dose (ADD) could be of particular
significance to the toxicologist in assessing the overall risk from
exposure to indoor formaldehyde. Depending on the specific details of
the toxicokinetics of formaldehyde the toxicologist may be interested in
understanding the exposure to humans on the worst (first) day of
exposure. The program allowed me to put in a single day (0.00274 years)
for the duration of the exposure but the results indicate that a
calculation error occurs for this input. The calculated ADC is above the
initial concentration and the time percentages come out to be >100%. If
this error is not easy or practical to fix then an explanation should be
included that advises the minimum time for duration of exposure.	The
model now has six default exposure groups covering different age ranges,
and the model results include year-by-year ADC calculations

Jayjock – Relative to alternate calculations methodologies it is
suggested that the modelers consider a modeling program based on The
Advanced Continuous Simulation Language (ACSL) computer language. ACSL
easily handles simultaneous differential equations while providing
parameter estimates from data via optimization and an intuitive
graphical interface.	The model as designed achieves the goal of
supporting EPA exposure assessments

Georgopoulos – The mathematical/algorithmic formulation and the
computational implementation of the model under review are adequately
explained, with sufficient clarity and level of detail; the
computational implementation of the model also appears robust and
accurate. Nevertheless, given the simplifying assumptions (discussed
further in the following paragraphs) in the formulation of the model,
and the significant variabilities inherent in the data sets that were
used to derive parameter values for FIAM-pwp, the level of confidence in
the indoor air concentration and exposure estimates provided by the
model, should in general be considered, at most, medium. In this
reviewer’s opinion, the uncertainty and variability levels in the
available data strongly suggest the need for a
distributional/probabilistic model formulation that explicitly employs
available data to develop ranges rather than point estimates of
concentrations and exposures.	The model as designed achieves the goal of
supporting EPA exposure assessments

Georgopoulos – Considering the (current) model formulation, FIAM-pwp
essentially consists of three components, i.e. (a) estimation of
formaldehyde emission rates from various types of pressed wood products
(based on data from various field and chamber studies), (b) estimation
of indoor mixing/dispersion through simple one-zone and two-zone models
that incorporate sink and back-pressure effects, and (c) simple
inhalation exposure calculations that employ average intakes for adults
and of children in different age groups. In this reviewer’s opinion, a
major drawback in the formulation of the FIAM-pwp is that components (a)
and (b) are not “separated” in a way that would allow, in a
straightforward manner, usage of the formaldehyde emission rates with
other available indoor air quality and inhalation exposure models. The
value of the information that has been incorporated in FIAM-pwp would be
greatly enhanced if its formulation explicitly considered decaying (e.g.
exponentially decreasing) emission rates and provided appropriate
formulas for implementing them (either within FIAM-pwp or in conjunction
with other, more comprehensive, indoor air quality and inhalation
exposure systems). Simple implementations of exponentially decaying
source with one-zone and two-zone indoor air models are well
established.	The model as designed achieves the goal of supporting EPA
exposure assessments



Unice – Evaluation of the decay curve for a single product or class of
products in isolation is likely to be more useful for understanding
chronic risk and informing policy decisions about managing chronic risk
than the suggested cross-sectional decay half-life approach. In
addition, new construction represents a minor fraction of total
residences, so the case where a new home is loaded with new products is
the exception rather than the typical case for most of the population.
In short, the emissions half-life is currently considered a ‘house
characteristic’, whereas I believe it would be more appropriate to
frame this in terms of a ‘product-specific’ characteristic.	EPA has
added clarifying text to  the revised documentation and/or help screen
text

Unice – Because the majority of the available reservoir is likely to
be depleted over the lifetime of the product, incremental risk from
separate products or collection of products can be added to derive total
risk or an acceptable risk range by type of product (e.g., MDF) could be
established. This incremental approach may slightly over-estimate true
risk depending on the assumed tenure of the occupants, but this can be
somewhat accounted for by assuming a finite emission lifetime of an
individual product (e.g., 10 years) or by selecting a target background
indoor concentration higher than ambient air levels corresponding to the
decay concentration where emissions from the source under consideration
would be considered de minimus.	EPA has added clarifying text to  the
revised documentation and/or help screen text

Unice – Under the current model guidance, the true increased cancer
risk attributable to installation of a specific product such as
wall-paneling in an existing home to repair damaged plaster or kitchen
cabinets installed during remodeling would not be reliably determined.
Initially, the backpressure effect will suppress a portion of the
formaldehyde emissions, such that if the user were to compare cumulative
dose with and without the product based on current model guidance, the
lifetime dose for that product will be underestimated. If the
equilibrium concentration (i.e., the theoretical concentration in the
absence of ventilation) of a product such as kitchen cabinets is less
than the pre-existing steady state concentration, running the model with
and without kitchen cabinets would result in the physically impossible
conclusion of negative incremental cancer risk because the cabinets
would initially act as a net sink.	EPA has added clarifying text to  the
revised documentation and/or help screen text

Unice – The assessment should be based on a conceptual model for decay
in formaldehyde emissions over time and basic research on the decay
profile of individual products should be encouraged in the model
documentation (rather than exclusively relying on the historical data
from the 1980’s). At a minimum the model should be revised to allow
consideration of periods of rapid and slow decay in the emission rate as
observed, for example, in Zinn et al. 1990, Mølhave et al. 1995, Myer
and Hermanns 1985 and Brown 1999.	EPA has added clarifying text to  the
revised documentation and/or help screen text

Unice – One conceptual model for assessing chronic exposure for
formaldehyde releasing products was published by Mølhave et al. (1995).
In this study, a selection of furniture and fitments supplied by a
retailer was placed into a large scale chamber and measurements were
made over an 8-week period. Conceptually, three exponential decay
compartments were attributed to three potential sources of formaldehyde
emissions, including:

• free formaldehyde (short-term compartment);

• decay of paraformaldehyde and other complex structures (medium-term
compartment); and

• hydrolysis of the resin (long-term compartment).

Using an exponential peeling procedure, the authors determined relative
sources strengths and half-lives for each of these three compartments. A
two-compartment (slow and fast decay) analysis using an EPA model also
adequately represented the transient data. The corresponding accumulated
dose was compared to the California Proposition 65 safe harbor level to
determine the incremental contribution of a specific set of products
(i.e., living room furnished with IKEA furniture). This product-specific
testing may be helpful in refining the estimate of LADD and LADC and the
approach to understanding whether compliance with Proposition 65 was
achieved was reasonable. I strongly recommend that the model
documentation be modified to make clear how this type of testing could
be incorporated into the model. Currently, the half-life parameter
pertains to a structure-based decay rate rather than a product-based
decay rate.	EPA has added clarifying text to  the revised documentation
and/or help screen text, as well as discussion regarding how the model
could be run with “fast” and “slow” decay periods

Unice – For the default case, the decay rate presented in Zinn et al.
(1990) seems more appropriate than the 2.92 year half-life derived from
cross-sectional sectional data in the original Versar model. Zinn et al.
(1990) estimated an apparent half-life of 216 days using approximately
one year of large chamber data for particle board and a natural
logarithm linear regression statistical analysis. The half-life reported
in Zinn et al. (1990) is not directly comparable to that adopted in the
original and current Versar model because Zinn et al. used the
relationship HCHO = F + G x ln(time) with regression constants F and G
whereas the EPA half-life was based on a traditional exponential decay
model. In lieu of the original logarithmic equation, the more generic
two-compartment (i.e., fast and slow decay) exponential decay model
could be fit the Zinn et al. data.	EPA has added clarifying text to  the
revised documentation and/or help screen text, as well as an discussion
regarding how the model could be run with “fast” and “slow”
decay periods



Unice – Finally, one concern is the interaction between ventilation
rate and decay rate. Given long-term testing data, the total reservoir
of formaldehyde and formaldehyde precursors for a given ventilation
condition could be calculated. Therefore, in the future when decay
characteristics are better understood, it should be possible to adjust
for ventilation rate and account for total mass of formaldehyde
generated over the lifetime of the product.	The model as designed
achieves the goal of supporting EPA exposure assessments

Unice – The original Versar model presented the linear regression
parameters both the Matthew’ model and the Hoetjer model (identified
as HBF model in 1986 Versar model), as well as the steady equations for
both models. These two models are both mass transfer based approaches,
however, the mass transfer coefficient obtained by linear regression is
not identical. This can be seen in the documentation for the 1986 Versar
model, where the correlation coefficient is typically higher for the
Hoetjer than for the Matthews model. Because slightly different results
are obtained depending on the regression model used and the Hoetjer
model is widely cited in the literature, it is recommended that the
model documentation be revised to include discussion of both approaches.
EPA has added clarifying text to  the revised documentation and/or help
screen text

Unice – The regression model for the HBF mass transfer relationship is
given in Myers and Nagaoka (1981). The linear regression models are:

where:  

K = -m = mass transfer coefficient (m/h)

CS = steady state concentration (mg/m3)

N = air change rate (1/hr)

L = loading rate (m2/m3)

Ceq = concentration at zero ventilation rate (N=0)

b = emission rate at zero formaldehyde concentration

The simplified steady state models are:

It can be shown these are identical models and that Ceq= b/-m.
Therefore, the difference in models is in the regression equation used
to derive the parameters, not the theoretical construct. The equilibrium
concentration is the vapor phase concentration at the surface of the
board and the bulk phase concentration that would be true in the absence
of ventilation. The Matthews parameter b is a combination of the mass
transfer coefficient K and equilibrium concentration Ceq. The mass
transfer coefficient is determined by the properties of the board or
object surface, whereas the equilibrium concentration is determined by
the formaldehyde generating properties of the panel and varies with
temperature, humidity and age (Berge et al. 1980). These concepts are
less transparently elucidated by the Matthews emissions equation. The
easiest way to incorporate a discussion of the two approaches would be
to show how m, b, Ceq and K are calculated for the kitchen cabinet data
presented in Koontz et al. (1996) using both the HBF and Matthews
regression equations (in an Appendix). The purpose of this appendix
would be to provide clear guidance on the type of chamber tests (i.e. at
least three N or L combinations for the linear regression) that would be
needed to determine the parameters that are input into the model.	EPA
has added clarifying text to  the revised documentation and/or help
screen text 

Unice – My preference would be to allow input the parameters from
either approach as was the case in the original model. Although the mass
transfer basis is identical for each model, the parameters of the
Hoetjer model are more intuitive. For example, based on the equilibrium
concentration, it is easy to see that the worst case concentration in
the absence of ventilation (N=0) is 0.19 mg/m3. It is easy to
conceptualize that this is approximately the level that would be
expected in a confined space like the inside of a closed kitchen
cabinet. The Matthews model parameters are less intuitive. The maximum
emission rate which can be derived directly from b is only true in the
absence of formaldehyde in the bulk air, or at the beginning of the
exposure period. However, the Ceq of the HBF model can be easily
calculated. The concentration in the absence of ventilation (and
background sources) is b/-m or Ceq= b/-m.	The model as designed achieves
the goal of supporting EPA exposure assessments

Unice – Having made the recommendation to allow for both approaches, I
recognize that is likely not possible due to time and financial
constraints – and given the essential identical model construct
perhaps of minimal benefit. My recommendation if only the Matthews model
is retained in the model is to provide an example calculation in the
model documentation showing how the user can start with the HBF linear
regression or the HBF model parameters and convert the parameters to the
equivalent Matthews model, or 

for the slightly different result that is obtained using the HBF versus
the Matthews linear regression models. Another option would be to
provide the HBF equilibrium concentration as supplemental output.
Knowledge about the equilibrium concentration allows an intuitive
determination about whether a particular product is likely to act as a
sink or a source based on an estimate of the anticipated indoor air
concentration. In conjunction with this conversion equation, a statement
should be included indicating that for plausible ranges of the N/L
ratio, selection of a particular linear regression model is not expected
to appreciably affect the results. In conjunction with this conversion
equation, a statement should be included indicating that for plausible
ranges of the N/L ratio, the selection of a particular linear regression
model is not expected to appreciably affect the results.	EPA has added
the equilibrium concentration as an intermediate output (on the Source
Screen)  and has revised the documentation and/or help screen text 

reduction in sophistication from that which is available in the prior
EPA/Versar MCCEM model.	Indoor sinks have been  removed from the model
because of  modeling uncertainties

Unice – As shown in Table 2, the presence of the carpet sink has a de
minimus effect on the net emission rate in Zone 1 given a source
generating a steady state formaldehyde concentration of 10 ppb.
Approximately 7 or 8 hours after introduction of the source, the rate of
increase in storage in the sink is negligible and therefore there is no
net effect attributable to the presence of this sink.	Indoor sinks have
been  removed from the model because of  modeling uncertainties

Unice – From this example, it should clear that when the Matthews
slope and intercept are assigned in accordance with the proposed
guidance, dM/dt is zero and there should be no net effect of the sink on
the emission rate. Any apparent effect is caused by minor deviations in
the intercept or rounding errors, which do not appear to have a
theoretical basis.	Indoor sinks have been  removed from the model
because of  modeling uncertainties

Unice – Some pressed wood formaldehyde “emitters” could
theoretically act as sinks under certain conditions based on the
properties of wood and the Matthews model (e.g. products with very low
equilibrium concentration or, equivalently, low Matthews model
intercept). For example, low emitting products with unreacted urea can
act as scavengers (Meyer and Hermanns 1985). This could potentially be
an interesting question for situations like manufactured home where a
product with excessive free formaldehyde levels could ‘contaminate’
otherwise low emitting products. Therefore, there may be some usefulness
in presenting the emission rates by product in the results, with
negative emission rate indicating a net sink effect. Because a steady
state model is used, one concern about implicit sinks, especially as
more low emitting products become available on the market, is that it is
difficult to account for whether or not the storage capacity of the
implicit sink has been exceeded over the time frame of interest. This
could be a minor potential concern in some situations, such as when few
high emitting products are paired with several very low emitting
products. Presentation of the emission rate as output would provide a
starting point for understanding whether implicit sinks are present.	EPA
has added the equilibrium concentration as an intermediate output (on
the Source Screen)  and has revised the documentation and/or help screen
text 

Unice – To provide an extreme example, the Mathew’s mass transfer
parameters for kitchen cabinets were paired with a range of assumed
background concentrations. This example illustrates two key points.
First, the use of the HBF model equilibrium concentration provides a
more intuitive way to anticipate the results than the Matthews intercept
b. Second, the output in the model of the emission rate attributed to
each of the individual components models could potentially provide
useful diagnostic information to the user or modeler. As wood panel
resin technologies improve and more ‘green resins’ become available,
the potential exists for low observed equilibrium concentrations.
Depending on the properties of the panel, these products may or may not
act as reversible and/or irreversible sinks. I believe the model should
provide sufficient ancillary information to understand the sensitive
parameters and identify other potential areas for testing.	EPA has added
the equilibrium concentration as an intermediate output (on the Source
Screen)  and has revised the documentation and/or help screen text 





2.1.2 Comment on the Default Scenarios and Assumptions Used by the
Model

Reviewer / Comment	Response

Jayjock – My suggestion is that the modelers should consider using and
including the seminal study by Murray and Burmaster as an important
reference for potential use with this model (ref: Murray and Burmaster:
Residential Air Exchange Rates in the United States: Empirical and
Estimated Parametric Distributions by Season and Climatic Region, Risk
Analysis, Vol 15, No. 4, 1995). Of particular value are data from
different geographical sectors and seasons.	EPA has added the suggested
reference to the revised documentation 

Unice – It is not clear how the manufacturer of a specific product
(e.g. MDF panel) could do a study with their specific product to refine
the cross-sectional residential half-life. Therefore, there is not clear
guidance regarding how to assess the degree to which some parameters are
conservative. I am concerned that too much emphasis is placed on the
large database of historical emissions parameters provided in the model
(Table 2-2) and historical decay rate evaluations (Section 3.6). I
believe the model formulation and model documentation should be more
forward looking in two respects. First, it should be made clear that
many of the default emissions parameters in the model are based on
historical testing and may or may not be representative of current
products. Second, clear guidance should be provided to end users (e.g.
importers, manufacturers, non-profit organizations or regulators) about
the data and steps required to assess the HCHO incremental exposure
attributable to the introduction of a specific product into the home.
EPA has added clarifying text to  the revised documentation and/or help
screen text, keeping in mind that incremental exposure is not the main
intent of the model



2.1.3 Comment on the Data Sources

Reviewer / Comment	Response

Jayjock – Note the above comment on residential air exchange rate and
a referenced source of good information on this important parameter. I
also do not know of other data sources on the emission of formaldehyde
from wood.	EPA has added the suggested reference to the revised
documentation

Georgopoulos – The data sources for FIAM-pwp appear reasonable.
However, since most of them appear be to rather old, one may question
whether they are still the most relevant data sources in relation to
currently available/used pressed wood products (given changes in
manufacturing procedures as well as in changes in relative levels of
domestically manufactured versus imported pressed wood products). The
answer may in fact be yes, but it should be explicitly stated and
supported.	EPA has added clarifying text to  the revised documentation
and/or help screen text

Unice – It is unknown the degree to which the data in Table 2-2
reflect manufacturing practices in 2009, and the potential for low or
high bias exists. Users of the model should be encouraged to collect and
use data for products currently in the supply chain. This could be
accomplished by deemphasizing the data in Table 2-2 (perhaps by moving
to Appendix), by requesting voluntary submission of chamber data from
industry prior to release of the model, or by having a mechanism whereby
the database in Table 2-2 is routinely updated as new information
becomes available.	EPA has moved Table 2-2 to an appendix and has added
clarifying text to  the revised documentation and/or help screen text

Unice – Airflow between the basement and outside was low, which may or
may not be representative for other structures. For example, if a lower
efficiency furnace is located in the basement, make-up air for
combustion could induce substantial airflow in the winter. To the casual
user, it will not be immediately obvious that Conventional 1 represents
the data for one test house under limited conditions. Therefore, I
recommend that this home be explicitly labeled in the model as the pilot
home to indicate that this data is for the home from the EPA pilot
study.	EPA has modified the house selections used in the model

Unice – More details regarding the analysis of PFT data used to derive
the airflow rates Conventional 2 should be provided.	EPA has modified
the house selections used in the model

Unice – The use of Langmuir isotherm for wallboard should be further
explained, as this was found to not be valid in the original report for
the pilot study (Koontz et al. 1996).	Indoor sinks have been  removed
from the model because of  modeling uncertainties (see Section 2.1.1)



2.1.4 Comment on Whether the Data Sources Provide Unused Information
That Would Assist in Quantifying Uncertainty or Variability for
Subsequent Model Use

Reviewer / Comment	Response

Zhang – Although the data sources are incomplete for quantifying the
uncertainty, it is possible to conduct an uncertainty analysis using
error propagation theory with assumptions on the uncertainties in the
raw data.	The model as designed achieves the goal of supporting EPA
exposure assessments

Jayjock – The Murray/Burmaster database on air exchange rates in U.S.
residences could assist in gauging the variability of this parameter.
Also, I believe the current data used in the model on emission sources
discloses a level of uncertainty and variability that could also be
mined by a person who is knowledgeable about these data and their
origins.	The model as designed achieves the goal of supporting EPA
exposure assessments

Georgopoulos – Yes: the variability in the available data could be
used to derive distributional estimates of emission rates that would
support probabilistic (and population based) indoor concentration and
corresponding exposure estimates with explicit characterization of
uncertainties and variabilities.	The model as designed achieves the goal
of supporting EPA exposure assessments

Unice – The approach to wallboard sinks used in the current EPA model
should be consistent with the wallboard sink equations and data
presented in Matthews et al. (1987). The use of Langmuir isotherm should
be further explained, as this was found to not be valid in the original
report (Koontz et al. 1996).	Indoor sinks have been  removed from the
model because of  modeling uncertainties (see Section 2.1.1)

Unice – The uncertainty and sensitivity analysis presented in Matthews
et al. 1985 could be discussed.	The Matthews analysis has been added to
an appendix in the revised documentation



2.1.5 Comment on Whether the Model Input Variables and the Estimates
Assigned to Them Appropriately Reflect Variability

Reviewer / Comment	Response

Jayjock – A user could choose to use inputted highest emission values
(no sinks) which maximize the concentration and exposure within the
range of his or her knowledge of the scenario to obtain a high end
prediction. This could be followed by a run in which he or she would use
inputs that were on the low end of the range (and include sinks) that
again reflected the modeler’s knowledge or confidence in the
situation. The comparison of outputs would provide some quantitative
level of variability extant within the system.	The model as designed
achieves the goal of supporting EPA exposure assessments

Jayjock – Another possibility would be to take the data in Table 2-2
and determine which data sets might be reasonably combined as presumably
coming from the same population. Parameters from these data sets could
be analyzed for the statistical variability and the parameters of the
emission equations could be represented as best-fit probabilistic
distribution functions.	The model as designed achieves the goal of
supporting EPA exposure assessments

Georgopoulos – As mentioned in the answers to previous questions, a
distributional/probabilistic approach, for defining model input
variables and deriving concentration and exposure estimates, would, in
this reviewer’s opinion, be more appropriate in relation to this
model.	The model as designed achieves the goal of supporting EPA
exposure assessments

Unice – The single exponential equation for decay is consistent with
the historical cross sectional residential data but inconsistent with
studies that have investigated single products or collection of products
over time (Zinn et al. 1990, Mølhave et al. 1995, Myer and Hermanns
1985 and Brown 1999). The true area under the curve for dose given a
long enough exposure time is likely better represented by the two- or
three- compartment exponential models than the single exponential model
because these models better represent the decay in the reservoir of
formaldehyde precursors in a specific product or collection of products.
EPA has added clarifying text to  the revised documentation and/or help
screen text, as well as discussion regarding how the model could be run
with “fast” and “slow” decay periods 

Unice – The mass transfer coefficient is not routinely measured in
most chamber testing. EPA should perform a sensitivity analysis to
determine whether measurement of the mass transfer coefficient is a data
need or data gap when newly acquired chamber test data is entered into
the model.	The model as designed achieves the goal of supporting EPA
exposure assessments

Unice – The model documentation does not provide a credible method for
selecting the indoor sink parameters.	Indoor sinks have been  removed
from the model because of  modeling uncertainties (see Section 2.1.1)

Unice – The model screen and documentation should make clear the
Conventional 1 represents a single unoccupied test home. The analysis
used to derive Conventional 2 should be described.	EPA has modified the
house selections used in the model

Unice – The degree to which the historical data in Table 2-2 reflects
domestic and imported production is not known.	EPA has moved Table 2-2
to an appendix and has added clarifying text to  the revised
documentation and/or help screen text



2.1.6 Comment on the Adequacy of the Documentation Supporting the
Overall Plausibility of the Exposure Estimates Generated by the Exposure
Model

Reviewer / Comment	Response

Jayjock – My sense is that the estimates have been clearly shown to be
plausible relative to results of comparing the modeled estimates to the
measured data in the Pilot House study. Even through all of the
practical difficulties, which were very well explained in the
documentation, this study appears to provide adequate evidence that the
model is certainly capable of providing credible estimates.	No
response/action required

Jayjock – Perhaps a more practical test of the model’s accuracy and
reliability might come from its comparison to measured data in some
realistic scenarios as viewed and implemented by or from the perspective
of a non-expert user.	EPA has added this type of comparison to the
evaluation of model results in Section 4.0

Georgopoulos – Ideally, the exposure estimates that are derived from
FIAM-pwp should be characterized within a total exposure context that
would take into account influences of numerous other indoor and outdoor
sources and sinks as well as of indoor and outdoor gas and heterogeneous
phase chemistry of formaldehyde (that may result in net production or
removal in a given microenvironment).	The model as designed achieves the
goal of supporting EPA exposure assessments

Georgopoulos – Comparison of model estimates with relevant
observational data sets for indoor-outdoor relationships and of personal
level (e.g. from studies such as RIOPA – see, e.g., the Liu et al.,
2006, reference in the Addendum of this review) would be valuable in
assessing, at least qualitatively, the plausibility and relevance of the
exposure estimates derived from FIAM-pwp.	EPA has added this type of
comparison to the evaluation of model results in Section 4.0

Unice – The EPA pilot study, while very useful, is limited to one
structure. I recommend pulling together a collection of representative
baseline loadings and comparing the model results to published
concentrations of formaldehyde in new and existing structures.	EPA has
added this type of comparison to the evaluation of model results in
Section 4.0

Unice – Average Daily Concentration/Dose EPA has not shown how use of
an overall residential half-life is to be used to derive
product-specific dose. Running the model with and without the product of
interest will not result in meaningful conclusions as discussed
previously. Accordingly, there is great uncertainty in product-specific
doses under the current model framework.	EPA has added an example
application and clarifying text to the revised documentation and/or help
screen text

Unice – EPA has not considered whether the historical 2.92 year
cross-sectional half-life is representative of the decay profile of
modern products or the potential variability in the formaldehyde
reservoirs and decay between products.	EPA has revised the default value
for the half life to 1.5 years and has added clarifying text to the
revised documentation and/or help screen text



2.1.7 Provide Any Other Types of Data to Assess the Reliability of the
Overall Exposure Model or Specific Model Elements

Reviewer / Comment	Response

Jayjock – Please see the discussion in 1.1 above in which an
additional critical emitter variable – Csat – is explicated using
the basic emission rate algorithm: Eqn 3-5. Data on this variable could
be directly measured in a straightforward manner. This could occur by
first constructing a box of the emitting material with an internal
volume sufficient to provide a good limit of analytical detection. The
box should be allowed to equilibrate such that the internal
concentration is no longer increasing. It is suggested that this would
occur within a week but could be confirm experimentally. After this
period, the exact internal air volume of the box should be sampled
quickly. The measured concentration represents 69% of the amount of
formaldehyde in the box if good mixing occurred during relatively rapid
sampling. (Note, having a small fan in the box will hasten equilibrium
and assure good mixing during sampling.)	EPA agrees that Csat is a
useful interpretation aid, but measurements such as those suggested here
are outside of the scope of this project

Georgopoulos – There are various studies and related data sets, both
from the US and abroad, that have appeared in recent years and could -
potentially - be used for evaluation of specific FIAM-pwp elements. The
Addendum to this review contains a selection of literature references
pertaining to such studies and data sets.	EPA has added this type of
comparison to the evaluation of model results in Section 4.0 ad selected
references have been added

Unice – As mentioned previously, I strongly believe a consensus
framework must be in place to assess product-specific long-term average
exposure. I believe the current lumped, or whole-residence based,
approach will create confusion and greatly complicate risk management
decisions if the end point of concern is based on chronic exposure.	The
model as designed achieves the goal of supporting EPA exposure
assessments



2.1.8 General Comments on Algorithms and Exposure Assumptions

Reviewer / Comment	Response

Tucker – Seems like a convoluted way to model emissions: The dependent
variable of primary interest – the HCHO concentration – is inherent
in the in the prediction algorithm; i.e., it is also used as an
independent variable. Nearly all the data appear to have been taken
under conditions where the HCHO concentration was well above 0.01 ppm,
which value is suggested by the model as a concentration of concern. So
the practical value of the slope in the emission rate equation is not
evident unless one is going to model situations that are clearly
unhealthy. Why not just use the intercept values and simplify the model?
EPA has carefully considered this comment but has not modified the
emissions algorithms

Tucker – That said, the analysis provided in section 5.3 of the
documentation suggests that whatever the convolutions involved, the
model gives respectable predictions. My reading of the tables in that
section, which summarize modeling results compared to measured values in
the EPA pilot study house, is that the model ranged from about 24% low
to about 65% high. Predictions within that range are quite good, given
the uncertainties in input data and unavoidable modeling assumptions.	No
response/action required

Tucker – Regarding the selection of a decay constant (or half-time if
you prefer) for predicting the rate at which the HCHO concentration will
decrease, I suspect the default choice of k = 0.23 yr-1 (or t1/2 = 3
years) might be improved by having different values for different time
periods. For example, the k value for the period 0-3 months may be
higher than the one for 3-12 months, which may be higher than for 12-24
months, etc.	EPA has added clarifying text to  the revised documentation
and/or help screen text, as well discussion regarding how the model
could be run with “fast” and “slow” decay periods

Tucker – One significant shortcoming of this model, and many indoor
models that assume complete mixing within compartments, is that it
doesn’t deal with what I refer to as “intimate” sources. Examples
are toys, cribs, bedding materials, clothing, and flooring materials for
children and bedding and clothing for elderly people. I think the most
straightforward approach to such sources is to assume a smaller
compartment or zone around such sources for the purpose of predicting
concentrations.	The model can be configured to address intimate sources,
by using a “virtual air space” around an individual exposed to an
intimate source as one of the model zones





APPENDIX B – DETAILED REVIEWER COMMENTS AND 

EPA RESPONSES: Part 2 – User Interface

2.2.1 Provide Comments on the Ease of Use of the FIAM-PWP Model

Reviewer / Comment	Response

Zhang – The interface design is easy to use in general.	No
response/action required

Zhang – The online version is very, very slow

	The  new version of the model runs on-line  under the IGEMs platform

Zhang – Berge and Myer - temperature effects. It is better to use one
only or combine it to provide a default number since users do not know
which one to use.	EPA has added clarifying text to  the revised
documentation and/or help screen text

Zhang – It would be useful to provide some sort of guide for users to
step through the modeling process (such as step 1, 2, 3, ….)	EPA has
added example applications to the revised documentation 

Jayjock – I found the user interface is remarkably clean and
straightforward.	No response/action required

Georgopoulos – Local use of the program in its current form would
still require setting it up so it is accessed by the browser – this
requires adjusting local computer settings in ways that may be
challenging for some systems – and users. Nevertheless, the web-based
version is sufficiently “user-friendly” and the data entry screens
are clear.	The  new version of the model runs on-line  under the IGEMs
platform

Georgopoulos – Maybe some “warnings” regarding the level of
uncertainties in the calculations could be incorporated, as reminders,
in the data-entry screens.	EPA has added clarifying text to  the
introductory screen for the model

Unice – I chose to use the local version of the software which
performed reasonably. I found the interface to be more cumbersome than
similar models like MCCEM, IAWX, EFAST or WPEM, but this is mostly
attributable to the use of a web-based platform and not a fault of the
developer.	The  new version of the model runs on-line  under the IGEMs
platform

Unice – Replacing this field (referring to air change rate) with plain
text would make clear that this value cannot be edited.	This field has
been “grayed out” to indicate that it cannot be directly edited

Unice – When a source is modified, the user should be required to
rename the source and a separate library of user entered sources should
be maintained. Currently, it is impossible to know whether someone
overrode the default values or the values presented are the default
values.	No longer applicable – library of sources has been changed 

Unice – Note also that the codes in the model do not match Table 2-2
in all instances.	The library of sources in the model has been changed 

Unice – When a screen is printed in standard portrait view, the edge
is cut off. It would extremely helpful to have a ‘Print All’ button
that would generate a single page for printing with all the inputs and
outputs suitable for printing in portrait view.	The  new version of the
model runs on-line  under the IGEMs platform

Unice – Please consider writing a routine to allow the same
information to be printed directly from the web-browser. (referring to
Excel output)	FIAM has been  revised accordingly, including
documentation and/or help screen text

Unice – In some instances the next box is out of order. (referring to
Tab key)	The  new version of the model runs on-line  under the IGEMs
platform

2.2.2 Comment on the Adequacy of the Model Help Screens

Reviewer / Comment	Response

Jayjock – In general, I found the content of the help screen to be
very good and, as expected, reflective of the documentation.	No
response/action required

Jayjock – My suggestion is to increase the length of the lines in the
help window to fill up the space and also increase the font size to use
up this additional utilized area.	The  new version of the model runs
on-line  under the IGEMs platform

Jayjock – I could not get the “audio help” to work.	The  new
version of the model runs on-line  under the IGEMs platform

Georgopoulos – The model-help screens appear generally adequate; they
could be enhanced by including some statements regarding the level of
precision expected for some of the entries.	EPA has added clarifying
text to  introductory screen for the model

Unice – The model help screens are intuitive and adequate. The audio
prompts generally worked well, although some were missing failed to
load.	The  new version of the model runs on-line  under the IGEMs
platform



2.2.3 General Comments on User Interface

Reviewer / Comment	Response

Tucker – I had little trouble with the user interface.	No
response/action required

Tucker – More careful attention to significant figures is needed.
FIAM has been  revised accordingly, including documentation and/or help
screen text

Tucker – On the Sinks Screen, it seems counter-intuitive that the B
value is positive rather than negative	Indoor sinks have been  removed
from the model because of  modeling uncertainties (see Section 2.1.1)





APPENDIX C – DETAILED REVIEWER COMMENTS AND

EPA RESPONSES: Part 3 – Documentation

2.3.1 Comment on Whether the References Have Been Accurately Identified

Reviewer / Comment	Response

Jayjock – The fourth reference (the CARB Final Regulation Order…)
should be identified as available online at:   HYPERLINK
"http://www.arb.ca.gov/regact/2007/compwood07/fro-final.pdf" 
http://www.arb.ca.gov/regact/2007/compwood07/fro-final.pdf  (last
accessed October 9, 2009). 

Similarly, the 8th reference (D. Hare…) can be found and should be
referenced at:   HYPERLINK
"http://www.ecobind.com/research/Evaluating_the_Contribution.pdf" 
http://www.ecobind.com/research/Evaluating_the_Contribution.pdf 

(last accessed October 9, 2009).	FIAM has been  revised accordingly,
including documentation and/or help screen text 

Jayjock – Indeed, any and all the reference particularly from the EPA,
GEOMET, Versar or industry groups should be identified by their web
links, if available, and if not available on the web, should be
considered for posting online with a link.	FIAM has been  revised
accordingly, including documentation and/or help screen text

Jayjock – The two additional references indicated above (Murray and
Burmaster on residential ventilation rate database and Jayjock on
“Backpressure” modeling) should be considered for inclusion.  
FIAM has been  revised accordingly, including documentation and/or help
screen text

Georgopoulos – The Addendum to this review presents a selection of
references. This reviewer would recommend consideration of at least some
of these references in future revisions of FIAM-pwp and its
documentation.	Selected references among those suggested have been added
to the revised documentation

Unice – The credit for the picture of the kitchen cabinets should be
provided.	No longer applicable – a new picture is being used that
requires no credit

Unice –For more information regarding formaldehyde emissions from
pressed wood products, model users could be referred to U.S EPA (1996)
ORD report – Sources and Factors Affecting Indoor Emissions from
Engineered Wood Products.	The reference has been added to the revised
model documentation

Unice – A basis should be provided for the recommend range of 1 to 10
ppb (referring to background concentration) on page 2-4. Similarly, it
should be explained why the concentration 10 ppb is ‘hard coded’
into the model in the percent of time > 0.01 ppm output.	FIAM has been 
revised accordingly, including documentation and/or help screen text

Unice – It may be helpful to provide citations to studies that have
summarized recent measured formaldehyde indoor air concentrations in new
and existing structures. Examples include Hodgson and Levin 2003 and
Hodgson et al. 2000. One of the concerns that these publications raise
is that the Matthews parameter database including in the model is
primarily for historical products, however, more recent indoor air data
shows that emissions have declined.	EPA has moved Table 2-2 to an
appendix and has added clarifying text to  the revised documentation
and/or help screen text, including some new references

Unice – Groah’s white papers and reports have been invaluable to me
in my past consulting work on formaldehyde. However, I recommend that
these references (essentially a white paper and report prepared on
behalf of HPVA) be replaced with peer-reviewed publications, conference
proceedings, or governmental citations. Examples of publications
addressing the long term decay of formaldehyde emissions from consumer
products include: Mølhave et al. 1995; Brown 1999; Zinn et al. 1990
(this study was relied upon by CARB); Myer and Hermanns 1985; and CARB
2007 – see Appendix B. One key point to be derived from these studies
is that a single exponential equation does not describe the decay in
emissions adequately.	EPA prefers to retain the  Groah reference and has
added discussion regarding how the model could be run with “fast”
and “slow” decay periods; selected references also have been added 

Unice – I recommend citing Jayjock 1994 in reference to the
backpressure effect first introduced on page 2-16 of the documentation.
EPA has added the suggested reference

2.3.2 Provide Comments on the Model Documentation (Does It Present the
Model Construct, Selected Model Inputs, and Model Results in a Clear,
Complete and Useful Manner?)

more logical to change the order of these chapters.	EPA prefers to keep
the original ordering of Chapters 2 and 3

Jayjock – In another relatively minor area I found the use of “D”
on page 3-4 to be unnecessarily complex.	EPA has removed the “D”
term from all equations

Georgopoulos – The documentation for FIAM-pwp is quite thorough in
presenting the model formulation and implementation as well as the data
sources that were used to develop parameter estimates for this
implementation.	No response/action required

Georgopoulos – However, the main issue with the documentation is the
justification of the formulation of certain elements of the model per se
(e.g., the assumption of constant rather than decreasing emission rates,
etc.) rather than of potential alternatives.	The model as designed
achieves the goal of supporting EPA exposure assessments

Unice – Consider explaining in the model documentation why the
concentration time points of 3, 6 and 12 months were selected for
output.	FIAM has been  revised accordingly, including documentation
and/or help screen text

Unice – As described in more detail earlier, the treatment of sinks is
not transparent and confusing because the rate of storage changes with
time but the model used is a steady state model. The approach used
appears to be more an attempt to modify the model results to reflect the
measured concentration than a rigorous handling of sinks. The approach
to wallboard sinks used in the current EPA model should be consistent
with the wallboard sink equations and data presented in Matthews et al.
(1987).	Indoor sinks have been  removed from the model because of 
modeling uncertainties (see Section 2.1.1)

Unice – Some stakeholders that may choose to use the new EPA
formaldehyde model may not be familiar with the linear regression
protocol required to estimate the parameters for either the Matthews and
Hoetjer formaldehyde mass transfer models. It is recommended that
documentation be revised to show an example linear regression for the
recent data Koontz et al. (1996).	FIAM has been  revised accordingly,
including documentation and/or help screen text

Unice – As described previously, the documentation should explain the
relationship between the Matthews and Hoetjer equation. The Hoetjer
equation or equivalent mass transfer equation is frequently cited in the
literature and some stakeholders may not be familiar with the Matthews
equation.	FIAM has been  revised accordingly, including documentation
and/or help screen text

Unice – The documentation should note that it is assumed that the
number of occupied adsorption sites are much less than the available
sites, reducing the isotherm equation to a simplified linear form.
Indoor sinks have been  removed from the model because of  modeling
uncertainties

Unice – Consider presenting the analysis of PFT measurements and raw
data in an appendix for Conventional 2 (p. 2-7). Is this data similar to
data available in MCCEM? What is the basis of 0.5 ACH for manufactured
home?	EPA has modified the house selections used in the model

Unice – Consider including in an appendix the analysis of the EPA
Pilot Study data used to determine the mass transfer coefficient and
intercept (p. 5-3).	EPA has added this type of analysis to an appendix
that contains historical chamber test data 



2.3.3 Provide Comments on Whether the Documentation for the FIAM - PWP
Makes Clear the Sources and Anticipated Magnitudes of Variability and
Uncertainty in Model Results

Reviewer / Comment	Response

Zhang – There are good description of the possible sources of
variability in the model prediction. However, it would be helpful if a
more quantitative analysis is conducted on the uncertainties of the
model coefficients and how they would propagate to the predicted
results.	The model as designed achieves the goal of supporting EPA
exposure assessments

Jayjock – I believe that the author does a particularly good job of
indicating the sources of relative uncertainty within the model itself
and within the basic data-sets and relationships that feed the tool;
however, I would have liked to see a level of quantitative uncertainty
analysis such as described above and herein. Specifically, the author
should consider presenting a range of outputs caused by a reasonably
anticipated range of input variables within all the particular
categories.	EPA has added an appendix on model sensitivity/uncertainty

Georgopoulos – Though extensive information on the data that were used
for the model is included, in this reviewer’s opinion, the types and
the extents of variability and uncertainty associated with various model
elements are not fully revealed in the documentation. An ideal solution
would be a probabilistic “re-formulation” of the model, with
explicit discussion in the documentation of the appropriate
distributions of values for the various inputs. As an alternative, even
with a deterministic formulation of the model, a table listing
explicitly various sources of uncertainty and variability associated
with specific processes and parameters of the model, along with a factor
or range specifying the approximate magnitude of each
uncertainty/variability, would be extremely useful.	The model as
designed achieves the goal of supporting EPA exposure assessments

Unice – Consider incorporating a mass transfer coefficient sensitivity
analysis into the model or model documentation. The uncertainty and
sensitivity analysis presented in Matthews et al. 1985 could be
discussed.	EPA has added an appendix on model sensitivity/uncertainty

Unice – Consider providing a database of homes similar to MCCEM
database to provide a larger range for interzonal airflows.	EPA has
modified the house selections used in the model; there are now five
generic structure types available

Unice – Consider including a simple Monte Carlo analysis as an
appendix to provide some guidance on whether refinement of a particular
input parameter is likely to change the conclusion.	EPA has added an
appendix on model sensitivity/uncertainty



2.3.4 Comment on Whether the Model Algorithms are Consistent with the
Descriptions in the Model Documentation

Reviewer / Comment	Response

Jayjock – Indeed, from what I can determine from studying this work
the model algorithms are for the most part consistent with their
descriptions in the model documentation. I did detect either an error in
sign or the omission of an explanation in the correlation coefficients
presented in Table 2-2.	FIAM has been  revised accordingly, including
documentation and/or help screen text

Jayjock – I would also prefer to see this “fit relationship”
described as the correlation coefficient squared (R2) because this value
represents the percentage of the variance in the model described by the
algorithm.	FIAM has been  revised accordingly, including documentation
and/or help screen text

Georgopoulos – The model algorithms appear consistent with model
documentation; issues with the algorithms were addressed in the answers
to questions of Part 1.	No response/action required

Unice – The model equations provided in the documentation were
reviewed. To further verify the model calculations were implemented as
intended using an alternate approach, an example loading was run to
steady state using IAQX Source Type 31. The model results were
successfully reproduced using this independent approach.	No
response/action required

Unice – The slope in the Matthews equation is not consistently
defined. For example on p 2-11, ER = mC+b, but on page 3-4, ER = -mC+b.
FIAM has been  revised accordingly, including documentation and/or help
screen text

Unice – One area that is particularly confusing is the model output of
average daily concentration (ADC) and lifetime average daily
concentration (LADC). From the model documentation it is not clear
whether the background concentration is included in the ADC and LADC.
EPA has added clarifying text to the revised documentation and/or help
screen text



2.3.5 General Comments on Documentation

Reviewer / Comment	Response

Tucker – I found the documentation to be generally clear and quite
useful. Users should have little trouble following it.	No
response/action required

Tucker – Page 2-4, 4th paragraph: you should not describe the slope as
“...in the form of a first-order sink representation...”	Indoor
sinks have been  removed from the model because of  modeling
uncertainties (see Section 2.1.1)

Tucker – Page 3-2, 2nd paragraph: the phrase “...flows of
formaldehyde mass into and out of a fixed volume must be equal...” is
not correct. In fact, equation 3.1 contradicts that.	EPA has added
clarifying text to the revised documentation and/or help screen text

Tucker – Page 3-4, first line: shouldn’t “independent” be
“dependent”?	EPA has added clarifying text to the revised
documentation and/or help screen text

Tucker –  While the term “half-life” has a certain value because
of its understandability, the normal way to determine it is from the
decay constant, which these days is a simple matter with spreadsheets
like Excel, where k can be determined easily from the “trendline”
equation for a first-order exponential decay of the concentration versus
time data.	EPA has added clarifying text to the revised documentation
and/or help screen text

Tucker – One area that I feel could be significantly improved is the
mathematical introduction to the model. I have sketched out an
alternative approach that I believe would be helpful to the general
user, and have attached it. I don’t necessarily suggest this as a
modification to FIAM, but as a general way to introduce mass balance
concepts.	EPA has carefully considered this comment but has not modified
the mathematical introduction



 T. Matthews, T. Reed, B. Tromberg, C. Daffron, and A. Hawthorne. 1983.
Formaldehyde Emissions from Combustion Sources and Solid Formaldehyde
Resin Containing Products: Potential Impact on Indoor Formaldehyde
Concentration and Possible Corrective Measures. Proceedings of ASHRAE
Symposium Management of Atmospheres in Tightly Enclosed Spaces, Santa
Barbara, CA. 

 T. Matthews, A. Hawthorne, and C Thompson. 1987. Formaldehyde Sorption
and Desorption Characteristics of Gypsum Wallboard. Environmental
Science & Technology 21: 629-634.

  Eastern Research Group, Inc. Peer Review Results for the Formaldehyde
Indoor Air Model – for Pressed Wood Products. Prepared for EPA Office
of Pollution Prevention and Toxics under Contract No. EP-W-05-014.

 USEPA 2012.  Formaldehyde from Composite Wood Products: Exposure
Assessment. Draft Final Report. U.S. Environmental Protection Agency,
Office of Pollution Prevention and Toxics, Exposure Assessment Branch,

1200 Pennsylvania Avenue NW, Washington, DC 20460.  July 2012.

 F. J. Offermann, J. Robertson, D. Springer, S. Brennan, and T. Woo.
2008. Window Usage, Ventilation, and Formaldehyde Concentrations in New
California Homes: Summer Field Sessions. ASHRAE IAQ 2007 Conference,
Baltimore, MD.

  PAGE   \* MERGEFORMAT  i 

  PAGE   \* MERGEFORMAT  15 

A-  PAGE   \* MERGEFORMAT  17 

B-  PAGE   \* MERGEFORMAT  2 

C-  PAGE   \* MERGEFORMAT  6 

Notice

Mention of the names of specific companies, organizations, or entities
does not constitute an endorsement by EPA.

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

 

 

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.1.1 Comment on the Adequacy of the Mathematical Equations …
(continued)

2.3.2 Provide Comments on the Model Documentation … (continued)

