Hostetler et al v. Johnson Controls Inc et al

Filing 406

OPINION AND ORDER GRANTING in part and DENYING in part Johnson Controls' Motion to exclude Dr. Keramida's opinions #393 . The Court excludes Dr. Keramida's calculations as to vapors from the plant and parking areas, and her upper-bound calculation as to TCE from the neighborhood groundwater. The Court otherwise denies the motion, however, so Dr. Keramida may present her calculations as modified to reflect the vapor concentrations from the neighborhood groundwater (with the site-specific attenuation factor) and the sewer pipes. Signed by Chief Judge Jon E DeGuilio on 10/8/2020. (Copy mailed to pro se party)(mrm)

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF INDIANA
SOUTH BEND DIVISION
AMOS HOSTETLER, et al.,
Plaintiffs,
v.
JOHNSON CONTROLS, INC., et al.,
Defendants.

)
)
)
)
)
)
)
)
)

Case No. 3:15-cv-226 JD

OPINION AND ORDER
Five plaintiffs assert claims against Johnson Controls, asserting that their homes have
been impacted by contamination from a former Johnson Controls facility. As relevant here, they
assert that the contamination has caused TCE and PCE vapors to enter the indoor air of their
homes. They claim that vapors in the soils at the site have migrated to their homes through the
soil and through sewer lines. They also allege that contaminated groundwater below their homes
produced vapors that migrated upwards to their homes. The Plaintiffs retained an expert, Dr.
Vasiliki Keramida, to calculate the concentrations of vapors that would have resulted inside their
homes through these various routes.
Johnson Controls moves to exclude those opinions under Rule 702, offering a litany of
criticisms of Dr. Keramida’s analysis. Some of its arguments reflect disagreements with her
inputs or conclusions, which are not grounds for excluding expert testimony. Others, however,
go to whether she reliably applied her methodologies to the facts of this case. As to those issues,
the Court agrees with Johnson Controls’ arguments, so it grants the motion in part.
A.

Dr. Keramida’s Analysis
Dr. Keramida employed a somewhat complex methodology to estimate the amount of

vapors that would have reached each plaintiff’s home. She analyzed four different sources that

could have contributed to the indoor air contamination: (1) vapors in the soil below the former
plant at the site (in the central and eastern portion of the site); (2) vapors in the soil below the
former parking area at the site (in the western portion of the site); (3) vapors from the
groundwater below the Plaintiffs’ homes; and (4) vapors from sewer lines. In the figure below,
the thick black line outlines the “plant area,” the thick blue and red lines to its left outline the
“parking area” for TCE and PCE, respectively. The yellow-shaded hexagons reflect the
Plaintiffs’ properties, and the thin colored lines reflect utility lines.

Vapors from the first three sources—the plant area, the parking area, and the
neighborhood groundwater—would have entered the homes (if at all) through the homes’ subslabs. For each of those three sources, Dr. Keramida calculated the amount of vapors that would
have traveled from those sources to each home’s sub-slab. She then added those concentrations
together to produce the total sub-slab vapor concentrations for each home. She then multiplied
that by a ratio of the indoor air and sub-slab vapor concentrations (the sub-slab to indoor air
attenuation factor) to determine the levels of vapors that the sub-slab vapors would have
produced in each home.

2

Dr. Keramida then calculated the concentration of vapors that would have entered from
sewer lines. To do so, she used the vapor levels that had been detected in sewer lines near each
home. She then multiplied that amount by an attenuation factor that she calculated to reflect the
ratio between sewer vapors and indoor air vapors. Finally, she added those amounts to the vapor
levels that would have entered through the sub-slab, producing the total vapor concentration in
each home, for each year of occupancy. She performed this same analysis for TCE and PCE, but
omitted the neighborhood groundwater as a source area for her PCE calculation, as PCE has not
been detected in the groundwater below the neighborhood.
Further complicating this analysis, Dr. Keramida used a different process for each source
area to determine the vapor levels those areas would have produced. For the plant area, Dr.
Keramida began her analysis with soil samples reflecting the concentration of contaminants in
the soil. From the soil samples, she calculated the vapor concentrations that the soil
contamination would have produced. She then averaged those concentrations, multiplied the
average concentration by the total volume of the site, and multiplied that by the “soil air filled
porosity” (reflecting the amount of air in the soil), to produce the total mass of vapor
contamination in this entire source area.
From there, she calculated how many of those vapors would have traveled laterally
through the soil and reached each home through the process of diffusion—movement from areas
of higher concentration to lower concentration. For that step, she used the “Crank Equation 3.5,”
a mathematical equation used to calculate what concentration will result a given distance away,
after a certain amount of time, if a certain mass is released from a certain point:

3

This equation assumes that a single mass is released from a single point, and that it diffuses into
an infinite volume. As the input for distance, Dr. Keramida used the distance between each home
and a point on the western edge of this source area—the closest point to the Plaintiffs’ homes
within the plant area. The output of that equation produced an estimated vapor concentration in
the sub-slab of each home for each year.
Dr. Keramida used a similar approach for the parking area, except that she used
groundwater samples instead of soil samples to calculate the total mass of contamination in that
area. From the groundwater samples, Dr. Keramida used an equation to determine the amount of
vapors that the groundwater contamination would have produced. She then averaged those
concentrations and, like for the plant area, multiplied that average by the total volume of the area
and by the soil air filled porosity, to arrive at a total mass of contaminants in this area. Dr.
Keramida then applied the Crank Equation using that mass as an input, to determine how much
vapor would have traveled by lateral diffusion from this area to each home. For the distance
input, she used the distance between each home and a point at the western edge of this area. The
result represented the estimated vapor levels at each home’s sub-slab attributable to the vapors in
this area.

4

Dr. Keramida used a different method for the next source—the vapors produced by
contamination in the shallow groundwater below each home. (Dr. Keramida did not use this step
for PCE, which has not been detected in the neighborhood groundwater.) Dr. Keramida first used
groundwater samples from near each home to estimate the amount of groundwater contamination
below each home. She then used an equation—the Johnson and Ettinger model—to determine
the levels of vapors that would volatize from the groundwater and migrate vertically to reach the
sub-slabs. This model can be used to evaluate the amount of vapors that will volatize from
groundwater, the diffusion through the soil to the sub-slab, and the transport across the building
slab and the vapors’ mixing with indoor air. Its inputs include parameters relating to the
groundwater and soil properties, chemical properties, and building properties. Dr. Keramida
applied this model to calculate the concentration of vapors at the sub-slab.
Having calculated the concentrations of sub-slab vapors that each of these three sources
would have produced, Dr. Keramida then estimated the levels of contamination those sub-slab
vapors would have produced in the indoor air. To do so, she multiplied the sum of those three
sources by two different attenuation factors. She first used a site-specific attenuation factor,
which she calculated by averaging the ratios of indoor air to sub-slab vapor concentrations from
samples taken in the neighborhood. Multiplying the attenuation factor by the sub-slab
concentration thus predicts the indoor air concentration. The site-specific attenuation factor Dr.
Keramida calculated was 0.003, meaning that if vapors were present in the sub-slab at 100
µg/m3, then 0.3 µg/m3 would be expected in the indoor air. Dr. Keramida also used the EPA’s
default sub-slab to indoor air attenuation factor, which represents an upper-bound level and is ten
times higher than the site-specific level, or 0.03. This step produced two alternative vapor levels
in each home attributable to vapor intrusion from the sub-slab.

5

Finally, Dr. Keramida calculated the amount of vapors that would be present in the
indoor air from the sewer gas. She calculated an attenuation factor by comparing the levels of
PCE in sewer samples and in indoor air samples of nearby homes. She also inspected each of the
homes to determine if they were susceptible to vapor intrusion from the sewer lines. She then
multiplied the attenuation factor by the average concentration of sewer gas samples taken in
sewer lines near each home, producing an estimate of the vapor levels that would have resulted
in each home from sewer gas contamination. She then added that amount to the amount of indoor
air vapors produced by sub-slab vapors, to produce her estimate for the total indoor air
concentrations that would have resulted in each home.
B.

Standard of Review
Rule 702 governs the admission of testimony by expert witnesses. Under that rule, a

witness “who is qualified as an expert by knowledge, skill, experience, training, or education”
may offer an opinion if the following criteria are met:
(a)

the expert’s scientific, technical, or other specialized knowledge will help
the trier of fact to understand the evidence or to determine a fact in issue;

(b)

the testimony is based on sufficient facts or data;

(c)

the testimony is the product of reliable principles and methods; and

(d)

the expert has reliably applied the principles and methods to the facts of the
case.

Fed. R. Evid. 702.
A court has a gatekeeping role to ensure that expert testimony meets these criteria.
Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579 (1993); C.W. ex rel. Wood v. Textron, Inc.,
807 F.3d 827, 834–35 (7th Cir. 2015). The proponent of the expert testimony bears the burden of
demonstrating that the testimony meets each of those elements. Varlen Corp. v. Liberty Mut. Ins.
Co., 924 F.3d 456, 459 (7th Cir. 2019). However, a court does not assess “‘the ultimate
6

correctness of the expert’s conclusions.’” Textron, 807 F.3d at 834 (quoting Schultz v. Akzo
Nobel Paints, LLC, 721 F.3d 426, 431 (7th Cir. 2013)). Rather, a court must focus “solely on
principles and methodology, not on the conclusions they generate.” Schultz, 721 F.3d at 432
(quoting Daubert, 509 U.S. at 595). “So long as the principles and methodology reflect reliable
scientific practice, ‘vigorous cross-examination, presentation of contrary evidence, and careful
instruction on the burden of proof are the traditional and appropriate means of attacking shaky
but admissible evidence.’” Id. (quoting Daubert, 509 U.S. at 596).
C.

Analysis
Johnson Controls moves to exclude Dr. Keramida’s opinions, asserting numerous

criticisms of her calculations. Johnson Controls argues that Dr. Keramida’s calculated vapor
levels are out of step with observed conditions (exceeding sampling results by over 1,000 times
in some instances), that she failed to fit her calculations to the facts of the case, and that she
failed to properly apply her methodologies to the inputs she selected, among other arguments.
The Plaintiffs disagree in each respect. More fundamentally, however, they also argue that these
arguments only reflect disagreements with Dr. Keramida’s inputs and conclusions or competing
opinions between experts, which are not grounds for excluding an opinion under Rule 702.
The crux of this dispute is how to characterize Johnson Controls’ objections: as criticisms
of Dr. Keramida’s inputs and conclusions, or arguments about whether she reliably applied her
methodology to the facts of the case. On the one hand, an expert must not only use a reliable
methodology, she must also “reliably appl[y] the principles and methods to the facts of the case.”
Fed. R. Evid. 702(d). On the other hand, Rule 702 authorizes courts to assess “the validity of the
methodology employed by an expert, not the quality of the data used in applying the
methodology or the conclusions produced.” Manpower, Inc. v. Ins. Co. of Penn., 732 F.3d 796,
806 (7th Cir. 2013); see also Stollings v. Ryobi Techs., Inc., 725 F.3d 753, 765 (7th Cir. 2013)
7

(“Rule 702’s requirement that the district judge determine that the expert used reliable methods
does not ordinarily extend to the reliability of the conclusions those methods produce—that is,
whether the conclusions are unimpeachable.”). “The soundness of the factual underpinnings of
the expert’s analysis and the correctness of the expert’s conclusions based on that analysis are
factual matters to be determined by the trier of fact, or, where appropriate, on summary
judgment.” Smith v. Ford Motor Co., 215 F.3d 713, 718 (7th Cir. 2000).
As the Seventh Circuit has recognized, “this is not always an easy line to draw.”
Manpower, 732 F.3d at 806. Whether a methodology produces accurate results will often depend
on the inputs an expert chooses to apply. However, “the selection of data inputs to employ in a
model is a question separate from the reliability of the methodology reflected in the model
itself.” Id. at 807. Thus, as the Seventh Circuit has summarized this point, “arguments about how
the selection of data inputs affect the merits of the conclusions produced by an accepted
methodology should normally be left to the jury.” Id. at 808; see also Stollings, 725 F.3d at 766–
67.
Some of Johnson Controls arguments plainly fall on the “inputs” or “conclusions” side of
that line. Others, however, take issue with whether she has reliably applied her calculations to the
facts of this case. In particular, the Court finds that Dr. Keramida has not shown that she reliably
applied her equation for diffusion from the plant and parking areas—which calculates diffusion
based on the release of a single mass from a single point—to the circumstances of this case, such
that it offers a relevant and reliable method of estimating the diffusion of vapors spread
throughout a multi-acre site. The Court also concludes that Dr. Keramida’s upper-bound
calculation of vapors through the sub-slab is unreliable because she failed to run her calculation
with the input she chose for that scenario. Johnson Controls’ criticisms of Dr. Keramida’s sewer

8

vapor calculations only address her inputs and conclusions, though, which are not grounds for
exclusion.
Before addressing Johnson Controls’ specific criticisms, the Court addresses its threshold
argument that Dr. Keramida simply assumed that vapors are in fact traveling through all of the
theoretically possible mechanisms of vapor transport. That’s not quite right. Dr. Keramida
considered various potential contributors to the vapor intrusion, but she did not just assume that
they were all occurring—the whole point of her calculations was to determine whether and in
what amount those sources would have contributed to vapor intrusion. For each source area, she
used sampling data to calculate the levels of contamination in that area. She then used equations
to estimate the extent to which that contamination would have reached each home. She also
calculated site-specific attenuation factors, which would have reflected the extent to which
vapors were migrating through the slabs or the sewer connections, at least on average. She also
inspected each of the homes to examine whether they were susceptible to vapor intrusion. [DE
393-2 p. 4–12, 23–25]. Johnson Controls can take issue with whether that analysis was reliable
and whether the conclusions were sound, which the Court turns to next, but Dr. Keramida did not
simply assume that vapor intrusion was occurring from each source.
1.

Plant and Parking Area Calculations

The Court begins with Dr. Keramida’s calculations of the vapors from the plant area and
parking area. For these source areas, Dr. Keramida used the “Crank Equation 3.5” to estimate the
amount of lateral diffusion from these areas to the Plaintiffs’ sub-slabs. Johnson Controls takes
issue with her application of that equation in multiple respects. Some of those criticisms plainly
take issue with her inputs, though. It argues, for example, that the input Dr. Keramida used for
the soil moisture was too low, and that the average depth to the water table is less than she

9

assumed. Those challenges to the inputs or factual underpinnings of her opinions are matters for
the jury to decide. See Smith, 215 F.3d at 718.
Other of Johnson Controls’ argument are more substantial, though. Most notably,
Johnson Controls argues that Dr. Keramida misapplied the Crank Equation by using a starting
point untethered to the facts of this case and not grounded in the equation. The Crank Equation
assumes that a single mass is released from a single point. The plant area is not a single point,
though, but a multi-acre site. 1 To apply this equation, then, Dr. Keramida needed to decide how
to adapt the equation to the fact of this case. That included deciding which point at the site to use
in order to make distance measurements for each home. In doing so, she “assumed that the point
source is located at the middle of the western edge of” each area. [DE 393-2 p. 89 (emphasis
added)]. In other words, she used a point at the border of the plant area closest to the Plaintiffs’
homes.
In the figure below, the point sources Dr. Keramida used for each area are marked in pen
with red x’s. The red x near the middle of the figure represents the point source for the entire
“plant area” to its right (shown by the thick black outline), and the red x to the left represents the
point source for the “parking area”:

1

Dr. Keramida used the same approach for the parking area, and the analysis is the same for that
area, but the Court refers to the plant area for simplicity.
10

[DE 398-14].
Thus constructed, the equation calculates the diffusion that would result if the entire mass
of contaminants throughout the plant area was collected and then released from that single point
closest to the Plaintiffs’ homes. Johnson Controls argues that this equation is not a reliable way
of calculating diffusion under the facts of this case, since vapors were actually spread across the
site, and the areas of highest concentration were near the central and eastern portions of the site,
far from the points on which Dr. Keramida based her equation. Johnson Controls argues that Dr.
Keramida artificially increased her projected vapor concentrations by orders of magnitude by
modeling the release from this point.
One way to characterize Johnson Controls’ argument might be as a criticism of Dr.
Keramida’s inputs—the distance input into the equation. The Court believes, however, that this
is better characterized as an issue of whether Dr. Keramida reliably applied a reliable
methodology to the facts of this case. The Seventh Circuit has held that disputes over the
appropriate inputs to use in a reliable methodology are not generally subject to scrutiny under
Rule 702. In those cases, though, the inputs in question have generally been factual issues that
11

the juries could have found to be correct, or discrepancies that juries could find did not affect the
applicability of the expert’s opinion. In Stollings, 725 F.3d 753, for example, one of the inputs
into the expert’s analysis was the effectiveness rate of a safety measure. The expert was not
asked to opine on that issue; the plaintiffs instead relied on another witness who testified that the
safety measure would work the “vast majority” of the time. Id. at 764. Based on that testimony,
the expert assumed an effectiveness rate of 90 percent. Though the court characterized that input
as “undoubtedly a rough estimate,” it held that the accuracy of this input did not go to the
reliability of the expert’s analysis based on that input, particularly since the expert’s opinion
would have been the same even if the rate was much lower. Id. at 766–67.
Likewise, in Manpower, 732 F.3d 796, a damages expert’s calculation depended in part
on the growth rate the expert used for the plaintiff’s revenues. The expert calculated the growth
rate based on a short window of time reflecting the plaintiff’s most recent performance. The
expert did so based on testimony that the company’s performance had improved beginning in
that period, making that period representative of its likely future performance. Id. at 801–02.
Though the district court found that such a short window could not reliably reflect the company’s
growth rate, the Seventh Circuit characterized that as a challenge to the factual underpinnings of
the expert’s analysis and the quality of its data inputs, which should not have factored into a Rule
702 ruling. Id. at 807.
In both those cases, there was evidence from which the jury could have found the inputs
or factual underpinnings to be accurate, whether through the expert’s testimony or through other
evidence. Because resolving those disputes is the jury’s job, a court cannot make its own factual
finding on those issues and exclude expert testimony as a result. A court’s inquiry is instead

12

whether the expert’s opinion would be relevant and reliable if the jury resolved those disputes in
the proponent’s favor.
By contrast, when an expert’s opinion is simply not grounded in the facts of the case, the
opinion will be excluded as unreliable and irrelevant. Owens v. Auxilium Pharms., Inc., 895 F.3d
971, 973 (7th Cir. 2018); see also Hartman v. EBSCO Indus., Inc., 758 F.3d 810, 819 (7th Cir.
2014) (stating that expert testimony must “fit the issue to which the expert is testifying and be
tied to the facts of the case”). In Owens, for example, a medical expert opined that the plaintiff
was injured by a drug he had been prescribed because the drug was capable of causing the
plaintiff’s injury when used as directed. The court excluded that testimony, however, as it was
undisputed that the plaintiff had not used the drug as directed. Even though the manner in which
the plaintiff used the drug might be described as a factual underpinning, the court explained that
the opinion’s exclusion was appropriate: “Of course, some questions regarding an expert’s use of
faulty assumptions or data should be left to a jury. But whether an expert’s approach lines up
with the basic facts of the case goes to the relevance and admissibility of the testimony itself.
Gatekeeping of this sort is properly left to the court.” 758 F.3d at 973. Likewise, when experts do
not offer a reliable method for moving from the factual underpinnings to their conclusions, the
opinions will not satisfy Rule 702’s reliability elements. See Gopalratnam v. Hewlett-Packard
Co., 877 F.3d 771, 784 (7th Cir. 2017) (“[The expert’s] reliability fails when it comes to the
method by which he derived conclusions from these underlying events.”); Bielskis v. Louisville
Ladder, Inc., 663 F.3d 887, 896 (7th Cir. 2011) (holding that the district court properly
determined “whether it was appropriate for the expert to rely on the test that he administered and
upon the sources of information which he employed”); Fed. R. Evid. 702(d).

13

Those latter circumstances better describe the issues here. There is no factual dispute over
where the Plaintiffs’ homes are located in relation to the site. The question is whether Dr.
Keramida has offered a reliable basis for bridging the gap between the existence of soil vapors at
that site and the amount of vapors that would result below those homes’ slabs. To do so, she
decided to use the “Crank Equation 3.5.” That equation calculates the extent of diffusion at a
particular distance if a single mass is released from a single point and then diffuses into an
infinite volume. There is no dispute that the equation reliably calculates diffusion under those
circumstances. The site here, however, is not a single point but a multi-acre property, and the
contamination is not concentrated in a single mass but is dispersed throughout the property in
varying amounts. Dr. Keramida thus has to show that she reliably adapted and applied the
equation to the conditions of this site such that its result is probative of the vapor levels at each
plaintiff’s home.
The Plaintiffs have not shown that she did so. Again, the equation calculates diffusion
based on the release of a single mass at a single point (not to be confused with mass contained
within an area or volume). For the mass input, Dr. Keramida calculated the total mass of vapors
spread all across each multi-acre area at the site. For the release points, she selected points on the
western border of each area. Thus, the equation Dr. Keramida ran calculated the diffusion that
would result if every molecule of contamination throughout the area was moved to the point
closest to the Plaintiffs’ homes and then released. (Johnson Controls likens this to the Big Bang
or to dropping a bomb on that point.) Neither the Plaintiffs nor Dr. Keramida have shown how
calculating the diffusion that would occur under those plainly counterfactual circumstances
offers a reliable basis for calculating the diffusion that would occur here, particularly given that
the vast majority of soil contamination is concentrated toward the center and eastern portions of

14

the site. Thus constructed, the equation does not fit the facts of this case or offer a reliable basis
for calculating diffusion of the contaminants at the site to the Plaintiffs’ homes.
The Plaintiffs argue that calculating diffusion from that point is appropriate to account for
the effect of the concrete slab on the ground level, which Dr. Keramida testified creates a “box”
that contains the vapors within the soil throughout the site. Their argument and Dr. Keramida’s
testimony, however, suggest that Dr. Keramida misunderstood and misapplied this equation.
There are other Crank equations that do measure the diffusion of a mass that is distributed within
an area or a volume. [DE 393-15 p. 166–69]. J. Crank, The Mathematics of Diffusion p. 29 (2d
1975). The equation Dr. Keramida relied on was not one of those, however, and her explanations
fail to show how her method reliably applies her calculation (based on a single mass being
released at a single point source) to the facts of this case.
Dr. Keramida testified, for example, that the concrete cover is important because it would
cause vapors to pool in the soil throughout the site instead of diffusing into the air:
But because of the cover, the vapors are collecting underneath. . . . So underneath
that cover then, . . . it’s like having a box, and all the vapors are going into that box;
and they commingle. So it is very appropriate to consider what is leaving the box
as a uniform average of what has gone into the box.
[DE 393-3 p. 26]. The equation Dr. Keramida applied, however, does not use a “uniform
average” of contamination across the plant area, but the total of all the mass throughout that area.
Likewise, Dr. Keramida’s calculation does not reflect vapors commingling throughout the multiacre site—the “box.” Instead, it calculates the effect of if every molecule of contamination
throughout the whole site was moved to the edge of the area closest to the Plaintiffs’ homes and
then released. Even under the “box” theory, vapors are not focused at a single point, as the
equation assumes, but are spread over multiple acres. Though some vapors would have exited the
“box” at the point Dr. Keramida used on the western edge, vapors would have equally exited
15

hundreds of yards away on the eastern edge, as well as along the northern and southern edges.
Her calculation, though, measures the effect of if the entire mass was released from a single point
closest to the Plaintiffs’ homes. 2
Dr. Keramida testified along those same lines that contaminants were “released through
the area” and that “they cover pretty much the entire site.” [DE 393-3 p. 20]. There is a
fundamental disconnect between that explanation—that contaminants were spread through the
whole site—and the calculation she actually ran—based on the entirety of the contamination
being released from a single point at the edge of the site. This explanation might make sense if
the equation was based on diffusion from a volume containing a distributed mass 3 [DE 393-15 p.
165–69], but the equation she chose instead uses a single, one-dimensional point source. Dr.
Keramida’s reference to the concrete slab falls short of explaining why calculating the effect of
releasing the entire mass at the edge closest to the Plaintiffs’ homes reliably reflects the diffusion
that would result from the presence of contaminants spread throughout the site.
In response to criticism by Johnson Controls’ expert that Dr. Keramida should have used
the center of mass at the site as the starting point for her equation, the Plaintiffs argue that Dr.
Dawson cited no authority for that opinion except for the Crank textbook, which is simply a
textbook of mathematical equations and says nothing about environmental sites. But it is the
Plaintiffs, as the proponents of this testimony, who have the burden of showing that Dr.
Keramida reliably applied a reliable methodology to the facts of this case. In that respect, their
argument that the Crank textbook is just a mathematical discourse that doesn’t describe

2

Calculating the effect of releasing higher levels of vapors at shorter distances from the
Plaintiffs’ homes naturally increases the resulting projections—by many orders of magnitude in
some instances, as noted below.
3
Dr. Keramida’s testimony that she used the point where “the vapors are exiting [the] source
area” suggests she applied her calculation as if it was such an equation.
16

environmental sites cuts the opposite direction. The Crank Equation may reliably calculate the
diffusion of a single mass instantaneously released from a single point into an infinite volume.
But Dr. Keramida has the burden of showing that she has reliably applied that equation to the
very different conditions of this multi-acre environmental site, such that her calculation is
relevant to this case. The Plaintiffs cannot meet that burden simply by criticizing the opposing
expert, and their own argument falls short in making an affirmative case for why Dr. Keramida’s
method is appropriate. See Zenith Elecs. Corp. v. WH-TV Broadcasting Corp., 395 F.3d 416, 419
(7th Cir. 2005) (“An expert must offer good reason to think that his approach produces an
accurate estimate using professional methods, and this estimate must be testable.”).
The Plaintiffs also argue in response that the equation includes other conservative
assumptions. The equation assumes, for example, a single instantaneous release instead of a
continuous release, which could produce higher values. It also assumes that the diffusion occurs
into an infinite volume in all directions, even though the concrete slab may have inhibited
vertical diffusion. 4 The problem, however, is that Dr. Keramida offered no analysis as to how
much effect those assumptions would have on the equation’s results or why that makes it
appropriate to model the release as if every bit of contamination was released from the edge of
the area. Dr. Keramida performed no calculations of how much vertical diffusion would have
occurred but for the slab or how the absence of diffusion in that direction would impact lateral
diffusion. She did nothing, in other words, to test the effect of these assumptions so that she
could reliably apply the calculations to the different circumstances of this case. A bare assertion

4

The Plaintiffs argue that Dr. Keramida used another conservative assumption by accounting for
the removal of contaminated soils in 1999. But even though Dr. Keramida testified that that
remediation removed TCE mass from the site, [DE 393-3 p. 33], her calculation reflected an
increase in total TCE mass. [DE 393-2 p. 78 (calculating an increase from 407 to 409 billion µg
in mass beginning in 2000)].
17

that the slab inhibits vertical diffusion does not suffice to explain why the manner in which Dr.
Keramida accounted for that effect is reliable. Zenith, 395 F.3d 419 (stating that an expert cannot
invoke “‘my expertise’ rather than analytic strategies”). Nor have the Plaintiffs cited any
literature or studies indicating that the Crank Equation can reliably calculate diffusion from an
environmental site if applied in this manner. Daubert, 509 U.S. 579, 593 (1993) (noting that
whether a theory has been tested or has been subjected to peer review and publication may bear
on the reliability of expert testimony).
Moreover, unlike in Stollings, where the input in question had no effect on the expert’s
opinion, the location of the mass’s release has a profound effect on Dr. Keramida’s calculation.
Even a 100-meter change in distance could change the vapor calculations by multiple orders of
magnitude. [DE 393-3 p. 48 (Table A-3), p. 79 (Table A-4), p. 89 (Table A-7); see also DE 3939 p. 57 (figure 17)]. That’s the difference between vapor levels in the tens of thousands versus
almost none at all. 5 Given the extreme variation caused by that parameter, the Plaintiffs would
need to offer some basis for concluding that the point source Dr. Keramida used is appropriate in
order for her calculation to be relevant and reliable when applied to this case, but the Court
cannot find that they have done so.
At bottom, Dr. Keramida’s use of the Crank Equation offers little more than speculation
wrapped in a fancy equation. That some of its inputs were founded on site data does not mean
that she reliably applied the methodology to the facts of this case in a way that makes the
calculation probative of the amount of vapors that would migrate from the source areas to any

5

For example, Dr. Keramida calculates the sub-slab vapors from the plant area for 1996 at 1109
Sander Ave to be 3.93 µg/m3, while her calculation for 1213 Egbert, just over 100 meters closer
to the point source, was 39,045.65 µg/m3—a difference of four orders of magnitude. [DE 393-2
p. 48].
18

home’s sub-slab. Of course, an equation’s parameters need not be perfectly aligned to the facts
of a case to be admissible; a calculation might still be relevant and reliable even if it carries a
degree of imprecision. The Court finds, however, that there is simply too great a gap between the
equation Dr. Keramida relied on and the actual conditions, and too little explanation and support
for whether she reliably adapted it to these facts, to find that the calculation is sufficiently
relevant and reliable under the circumstances of this case.
For those reasons, the Court excludes Dr. Keramida’s opinions as to the contributions of
TCE and PCE from the plant and parking areas. Having excluded the opinions on that basis, the
Court need not also resolve whether Dr. Keramida’s use of a simple average instead of a
weighted average was a reliable method for calculating the total mass in the plant area, when the
samples were not taken at equal or random intervals across the site but were focused most
heavily in the areas of greatest contamination. 6
2.

Upper-Bound TCE Calculations

Johnson Controls next takes issue with Dr. Keramida’s upper-bound calculation for TCE.
Recall that Dr. Keramida used two different attenuation factors to estimate the extent to which
the sub-slab vapors would impact the indoor air: a site-specific attenuation factor (0.003) and the
6

The Court notes, however, that the Plaintiffs’ arguments on that topic reflect misunderstandings
of both sides’ experts’ opinions in this respect, and also mistake concentrations for mass. For
example, the Plaintiffs assert that the total TCE mass in a given area is the sum of all samples
taken in that area. [DE 398 p. 17]. But the samples reflected only concentrations (µg/kg)—
ratios—not mass (µg). To convert concentrations to mass, the concentrations have to be
multiplied by the mass or volume of the medium containing the contaminant in order to
determine the mass of the contaminant. Thus, both sides’ experts first averaged the
concentrations before multiplying them by the volume to calculate mass. [See DE 393-2 p. 78
(multiplying the average concentration (µg/m3) by the volume (m3) to calculate mass (µg))]. Dr.
Dawson just did so within smaller areas instead of across one large area. The Plaintiffs likewise
assert that Dr. Dawson calculated the TCE mass in a particular cell as 71,993 µg/kg. As reflected
by the unit, though, that is a concentration, not a mass. [DE 398-10 p. 6]. To calculate the mass
within that cell, Dr. Dawson multiplied that average concentration by the volume of air in that
cell. [DE 393-9 p. 53].
19

EPA’s default upper-bound attenuation factor (0.03). She multiplied those attenuation factors by
her predicted sub-slab vapor levels, to estimate the indoor air vapor levels. Johnson Controls
argues that Dr. Keramida failed to recognize, however, that the sub-slab to indoor air attenuation
factor is also an input in the Johnson and Ettinger model, which she used to estimate the sub-slab
vapors produced by TCE in the groundwater below each home. Because Dr. Keramida did not
run the Johnson and Ettinger model using her upper-bound attenuation factor, Johnson Controls
argues that Dr. Keramida failed to reliably apply her methodology to the inputs she chose. The
Court agrees.
The sub-slab to indoor air attenuation factor is a comparison between the concentrations
of vapors below the slab and in indoor air. To slightly oversimplify, this attenuation factor
reflects in part how easily the vapors travel through the slab into the indoor air. A higher
attenuation factor means that more vapors are traveling through the slab into the indoor air,
resulting in higher concentrations of vapors in indoor air. [DE 393-11 p. 4]. This also means,
however, that because more vapors travel through the slab, fewer vapors pool below the slab. See
id. p. 8. Thus, using the Johnson and Ettinger model, a higher attenuation factor results in lower
levels of vapors in the sub-slab, id.; more vapors are traveling through the slab instead of pooling
below the slab.
Dr. Keramida’s calculation assumed a low sub-slab to indoor air attenuation factor
(0.003) when using the Johnson and Ettinger model to calculate the concentration of vapors
below the slab. This low attenuation factor resulted in higher levels of sub-slab vapors. To make
her upper-bound calculations of TCE in indoor air, however, Dr. Keramida used an attenuation
factor ten times higher (0.03). With that higher attenuation factor, more of the sub-slab vapors
enter indoor air but fewer vapors pool below the slab, so the Johnson and Ettinger model

20

produces a lower amount of vapors at the sub-slab (as explained here by Dr. Ettinger, co-author
of the model [DE 393-11 p. 7–10]). Yet, Dr. Keramida did not re-run her calculation at the
previous step to calculate the level of vapors that would result below the slab using that
attenuation factor. Surprisingly, Dr. Keramida failed to recognize that her assumption at this step
was also an input at the previous step and would change (substantially reduce) the levels of
vapors in the sub-slab. She thus did not re-run the Johnson and Ettinger model with this new
attenuation factor, but instead used the higher sub-slab vapor levels that would be produced by a
lower attenuation factor.
In other words, Dr. Keramida’s upper-bound calculation rests on two contradictory
assumptions: that a low attenuation factor is causing vapors to pool below the building slabs in
greater amounts, and that a high attenuation factor is causing more of those vapors to enter the
indoor air from below the slabs. The problem is not, as the Plaintiffs attempt to characterize the
argument, whether she should have used a particular attenuation factor. Rather, the problem is
that, having chosen a given attenuation factor for her upper-bound calculation, Dr. Keramida did
not reliably apply her own calculations using that input. [DE 401 p. 11 (“[T]he issue is that,
having chosen that input, Keramida was obligated to correctly apply it within the J&E model.
She did not.”)]. If the sub-slab-to-indoor-air attenuation factor she used for the upper-bound
calculation is correct, then she did not reliably apply her methodology at the previous step to
estimate the amount of vapors that would be present below the buildings. This disconnect is fatal
to the reliability of her opinion on the upper-bound levels of TCE in indoor air. See Stollings,
725 F.3d at 766 (stating that an expert’s testimony must be “based on a correct application of a
reliable methodology” (emphasis added)).

21

The Court thus excludes Dr. Keramida’s upper-bound opinion as to TCE from the
neighborhood groundwater. This does not affect the groundwater calculations Dr. Keramida used
based on an attenuation factor of 0.003, though, since that is the same input she used for the
Johnson and Ettinger model. Though Johnson Controls disagrees with Dr. Keramida’s opinions
on that issue, it does not offer any specific objections to her methodology for her lower-bound
calculations. Dr. Keramida can thus offer her calculations as to the TCE vapor levels that would
result from the neighborhood groundwater based on that attenuation factor.
3.

Sewer Vapor Calculations

Last, Johnson Controls moves to exclude Dr. Keramida’s opinions as to vapor intrusion
from the sewer lines. As to these issues, the Court finds that Johnson Controls’ arguments go
only to Dr. Keramida’s inputs, selection of data, and conclusions, rather than the reliability of her
methods, so the Court denies the motion in this respect. To calculate the amount of vapor
intrusion through the sewers, Dr. Keramida began with sampling data reflecting the levels of
TCE and PCE vapors in sewer lines near the homes. To determine what effect those sewer
vapors had on indoor air, she calculated an attenuation factor by comparing sewer samples and
nearby indoor air samples. She used samples from eight different homes and averaged the ratios
together to produce a site-specific attenuation factor. Dr. Keramida explained that she used PCE
samples for that step because PCE was detected in almost all of the samples while TCE was not,
so using PCE allowed for more comparisons. 7 Then, for each home, Dr. Keramida took an
average of the sewer vapors detected in nearby sewer lines and assumed that those levels
7

Johnson Controls argues in a footnote that Dr. Keramida double-counted the indoor air vapors
by assuming for the purposes of this attenuation factor that all of the vapors came from the
sewers, while assuming for her sub-slab attenuation factor that all vapors came from the subslab. The Court has excluded Dr. Keramida’s opinions as to PCE entering from the sub-slab,
though, and Johnson Controls argues that no PCE was entering the indoor air through the subslab.
22

represented an average over time. Finally, she multiplied those levels by the attenuation factor
she calculated, producing her estimate for the vapor levels in each home attributable to the
sewers.
Johnson Controls takes issue with Dr. Keramida’s calculation of the attenuation factor. It
first disagrees with her decision to calculate the attenuation factor based on the PCE samples
instead of TCE (which would have produced a much lower attenuation factor). This is a
disagreement with an expert’s selection of data, though, which is not grounds for exclusion.
Manpower, 732 F.3d at 809 (“That the reasoning behind that choice [of data to use] could be
challenged as incomplete or faulty does not make it any less grounded in real data.”). Johnson
Controls does not dispute that comparing sewer vapor samples to indoor air samples is an
appropriate way of determining an attenuation factor. 8 Dr. Keramida did so here, using samples
of PCE. Her use of those samples instead of TCE samples may be grounds for cross-examination
or for offering contrary evidence, but does not undermine the reliability of her method.
Johnson Controls also objects that Dr. Keramida applied the same PCE-based attenuation
factor to her projections for both TCE and PCE in indoor air. Yet it argues at the same time that
the attenuation rates for TCE and PCE should be the same. Thus, her use of a site-specific
attenuation rate to calculate TCE vapors is not unreliable just because it was derived from PCE
samples. Johnson Controls can argue at trial that the TCE levels from the same sampling results
would produce a much lower attenuation factor, and that the PCE in indoor air might have been
due to other sources. But those would be matters of weight, they do not affect whether Dr.
Keramida applied a reliable methodology.

8

Johnson Controls asserts that Dr. Keramida did not consider building ventilation, but the
attenuation factor inherently accounts for building ventilation.
23

Johnson Controls also argues that the attenuation factor Dr. Keramida calculated is
“exceedingly improbable.” That’s a challenge to her conclusion at this step, though, which is
likewise misplaced in a Rule 702 analysis. Johnson Controls further argues that some of the
inputs Dr. Keramida used had implausibly high ratios of indoor-air to sewer-gas levels, given the
ventilation rates of indoor air and the dilution that would occur when sewer gases mix with
indoor air. As already explained, though, that is a challenge to Dr. Keramida’s selection of data
to rely on. Johnson Controls is free to argue that the attenuation factor Dr. Keramida calculated is
implausible and based on faulty data. It can also disagree with her explanations for whether her
results are consistent with literature and whether she has plausibly explained away the difference
between the PCE and TCE levels. But Dr. Keramida applied an appropriate methodology in
calculating the attenuation factor, which suffices for Rule 702. The Court thus denies the motion
to exclude Dr. Keramida’s projections as to the vapor levels from sewer gases.
D.

Conclusion
The Court grants in part and denies in part Johnson Controls’ motion to exclude Dr.

Keramida’s opinions. [DE 393]. The Court excludes Dr. Keramida’s calculations as to vapors
from the plant and parking areas, and her upper-bound calculation as to TCE from the
neighborhood groundwater. The Court otherwise denies the motion, however, so Dr. Keramida
may present her calculations as modified to reflect the vapor concentrations from the
neighborhood groundwater (with the site-specific attenuation factor) and the sewer pipes.
SO ORDERED.
ENTERED: October 8, 2020
/s/ JON E. DEGUILIO
Chief Judge
United States District Court

24