it
THE EPA RADON MITIGATION TEST MATRIX:
FRAMEWORK AND INITIAL PRIORITIZATION EFFORTS
Prepared by
Air and Energy Engineering Research Laboratory
Office of Environmental Engineering and Technology
Office of Research and Development
U. S. Environmental Protection Agency
August 1986
-------
Table of Contents
Subject
Background
Objectives
Key Features
i
Independent Variables
Matrix Format
Example of Nij Calculation
Status
Preliminary Effect of Prioritize
Matrix Cells
Technical Issues Needing Review
Tables
Subject
1. Range of Variables (Existing Houses)
2. Range of Variables (New Construction)
1. Matrix Format
4. Approach (Example of Nij Calculation)
5. House Estimates
Page
1
2
3
6
11
13
20
21
26
Page
8
10
12
14
.24
-------
THE EPA RADON MITIGATION TEST MATRIX:
FRAMEWORK AND INITIAL PRIORITIZATION EFFORTS
Background
As part of the Agency effort to address the issue of indoor radon, the
Office of Research and Development is conducting & program to develop and
demonstrate cost-effective radon reduction measures for single-family houses
(including existing houses, and new houses under construction). This program
is national in scope,1 addressing the full range of residential substructure
types, radon levels and geological/meteorological conditions representative of
the eatire country.
In order to assure effective coverage of the wide range of variables
needed for a nationally-representataive program, a test matrix is being developed.
This matrix will define the number of existing and new houses of each
substructure type, that would have to be tested with each radon mitigation
technigue to achieve a given degree of statistical confidence, considering the
other variables of importance in the design and performance of the mitigation
system. This matrix will serve, not only as a guide to avoid duplication and
omissions in ORD's own testing, but as a mechanism by which installations
made by others—such as installations resulting from ORP's House Evaluation
Program—might contribute to satisfying segments of the overall data requirements.
This document is a description of the initial efforts to develop this matrix.
Specifically, this document outlines the framework within which the matrix will
be developed; and it describes some initial efforts to prioritize the elements
in the matrix. As discussed later, such prioritization is important because
the number of houses that would have to be tested in order to address all of
-------
the variables of concern, 'is too large for all variables to be addressed in a
short time.
Efforts to develop the matrix, within the framework described here, are
underway. But even after the matrix is initially defined, it is expected that it
will be modified and updated regularly as further information becomes available
during the course of the Radon Mitigation Demonstration Program.
Objectives
The overall objective of the test matrix is to define a radon mitigation
field testing program which will:
1. be technically defensible
2. most efficiently and quickly put the Agency in a position to suggest
cost-effective radon mitigation alternatives.
3. provide known confidence in the performance of these mitigation
alternatives, to reduce the risk that techniques might not perform
as expected in an application for which the Agency suggested them.
4. initially focus on the house types and the other particular conditions
(e.g., house design details, geological conditions) which are responsible
for a) the greatest cumulative population exposure nationwide, and b) the
most acute individual exposure.
5. ultimately provide mitigation alternatives for any homeowner in the U.S.
under any conditions.
The priority concern with acute exposures, in 4 above, naturally results
from the dramatically increased lung cancer risks that occupants of worst-case
houses face. Thejconcern with cumulative exposure results from the possibility
that very large numbers of people exposed to relatively low radon levels might
have a greater combined dosage than the smaller numbers of people who live in
;
"hot spots."
- 2 -
-------
The ultimate objective of assuring the availability of alternative
technologies applicable for any homeowner under any conditions, is ambitious.
..Not all conditions can be investigated immediately, due to limitations in
available expertise and resources. It is for that reason that the initial
prioritization of the matrix--to determine conditions responsible for the
greatest acute and cumulative .exposures--is so important, to direct near-
( *
term testing.
As a corollary objective, the test matrix is intended to provide a
technical basis for selecting test sites from among those which become avail-
able. It will probably not be feasible to conduct EPA-sponsored mitigation
testing in every state where an indoor radon problem is discovered. The
matrix will assist EPA in making rational selection of those candidate
sites which would produce the most required priority data.
Key Features • • •
The major function of the test matrix is to define the number of houses,
of each substructure type, in which each individual mitigation measure should
be tested under each set of conditions. A "set of conditions", as considered
here, refers to a set of selected values for the different independent variables
r
(discussed in following section) which might influence the design and performance
of the mitigation systems. In developing these numbers under the matrix, a
combination of engineering judgement and statistics will be utilized in an
effort to address, as efficiently as possible, those independent variables
recognized now as possibly being of practical importance. The basic intent
of the matrix approach is to define the performance (and the necessary design
features) of the various techniques under each applicable set of conditions, to
- 3 -
-------
a pre-selected degree of statistical confidence, through testing in the
minimum number of houses. Prioritization of the matrix will be attempted
m
in£jsn effort to assure where feasible that the most important sets of condi-
ti« is are addressed,first.
ho
ho
of
ti
ho
in
ui
ad
if
of
tht
1
There will, in fact, be two matrices: one addressing testing on existing
ies, and a separate one addressing mitigation features incorporated into new
tes under construction.
The basic approach of the matrix is to initially include a limited amount
esting at each set of conditions which is of interest. At the present
, "limited" is defined as five houses. The data from these first five
es would be analyzed to define the data variability (the confidence
rvals) at that set of conditions. Based upon these initial results, the
er of houses could then be increased (or the testing redirected to
ess other variables which might be responsible for the observed variability)
ecessary to narrow the confidence interval to the desired degree.
i-
An underlying philosophy of the matrix approach is to minimize the amount
esting that will be required to achieve the goals. Efforts to minimize
'testing include:-
1. conduct of the testing in an incremental,, step-by-step manner. As
described in:the previous paragraph, 1imited tests are conducted
first, and decisions then made.prior to further testing at a given
set of conditions. In this manner, if testing needs to be redirected,
such redirection can be accomplished before extensive testing is done;
and once the confidence interval is narrowed to the goal level, testing
at that set of-conditions can be stopped. .
- 4 -
-------
2. use of a fractional factorial experimental design, rather than a
full factorial. prhe fractional method reduces the number of houses
that must be tested to one-half or one-quarter the number that would
be required by a full factorial, in some cases. This reduction is
accomplished by testing only one-half to one-quarter of the possible
sets of conditions, then using statistics to separate out the individual
effects of each variable.
3. utilization of data from other investigators. Data from other sources
will be used to complete portions of the matrix wherever the data
quality from the outside sources is known and is adequate.
The number of houses included for testing in the matrix will be subdivided
according to whether or not detailed diagnostic testing will be performed.
All of the houses tested will include some diagnostic besting to help
\,
understand radon,entry routes and why an installed system is or is not performing
well; such diagnostic testing could include, e.g., spot radon measurements at
specific locations within the house, pressure and flow measurements in the
piping associated with the mitigation system, etc. But the houses involving
detailed diagosti'cs--perhaps 15% of the total houses—will have more compre-
hensive pre- and post-mitigation testing, involving a greater array of
diagnostic techniques (and a higher cost per house), with the intention of
gaining a more fundamental understanding of the house dynamics and of mitiga-
tion system performance. It is felt that the most cost-effective program
will have a suitable balance between houses involving detailed diagnostics
(to improve fundamental understanding, and to thus help guide the remainder
of the program), and houses involving less detailed diagnostics (to
develop, at reasonable cost, a data base sufficiently large to permit
- 5 -
-------
'dequate statistical analysis). One other product of the detailed diagnostic
testing will be protocols for house diagnosis and diagnostic testing.
Such protocols can be utilized by others/ so that other radon diagnosticians
can do the job in a complete and consistent manner, and so that the data
of others can most effectively be employed by EPA to satisfy segments of
the data requirements under the matrix.
The matrix addresses only that part of the radon mitigation development/
demonstration program involving field testing in houses. The field testing,
of course, is the major element of the mitigation program. Other elements
include: laboratory studies of specific mitigation approaches preparatory to,
i
or in support of, field work (e.g., laboratory testing of air cleaner performance
or of sealant performance/durability); and technology transfer activities (such
s preparation of the mitigation brochure and manual).
Independent Variables
Six categories of independent variables are currently being considered in
the design of the test matrix:
1. House substructure type (e.g., basement versus slab on grade versus
crawl space),
2. Mitigation technique
3. House construction details within a given substructure type (e.g.,
whether or not a fireplace is present)
4. Initial radon concentration
5. Geological and meteorological conditions, insofar as they might
influence mitigation performance
6. Mitigation technique design and operating conditions, within a given type
of mitigation technique (e.g., whether a given active soil ventilation
technique is operated to draw suction or to pressurize).
- 6 -
-------
The definition of the specific variables of importance within each of these
categories is still on-going. Between three and fourteen variables have been
identified to date within the different categories; generally, there are at
least two levels (possible values) for a specific variable.
The specific variables within each category for the matrix addressing
existing houses are listed in Table 1, insofar as the variables have been
defined to date.
Not all combinations of these variables shown in Table 1 need to be
considered. For example, under the category of "House Design Details", the
number of levels in the house (one story versus two story) could be very
important if the mitigation technique being considered were a heat recovery
ventilator, but the number of levels would not be expected to be important
if the mitigation technique were sub-slab ventilation. As another example,
under the category of "Initial Radon Concentrations", only the low and inter-
mediate initial concentrations would be considered for heat recovery venti-
lators, since the degrees of reduction normally achievable with that miti-
gation technique are not sufficiently high for those devices to be applicable,
by themselves, to high initial concentrations.
Table 2 lists the specific variables within each applicable category
for the matrix addressing new houses in the design/construction stage. The
number of variables (and categories) for the new house matrix are currently
more limited than for the existing house matrix. Basically, this occurs
because the mitigation approaches can be generally limited to closing off
soil gas entry routes during construction (perhaps in combination with
installation of passive or active ventilation systems). These construction
- 7 -
-------
TABLE 1
RADON MITIGATION TEST MATRIX
RANGES OF VARIABLES (EXISTING HOUSES)
,u
BEING
HOUSE SUBSTRUCTURE TYPE
-
I
.
lAL
JASEMENT - HOLLOW BLOCK FOUNDATION WALLS
ASEMENT - POURED CONCRETE FOUNDATION WALLS
ASEMENT - F1ELDSTONE FOUNDATION WALLS
jj> SLAB ON GRADE
5. CRAWL SPACE
COMBINATIONS OF THE ABOVE ARE ALSO POSSIBLE, BUT ARE NOT
IDENTIFIED ON THE MATRIX'
MITIGATION TECHNIQUE
HOUSE VENTILATION
1. HEAT RECOVERY VENTILATORS (AIR-TO-AIR HEAT EXCHANGERS)
2. NATURAL VENTILATION
3. FORCED VENTILATION
SEALING
4- COMPREHENSIVE SEALING
ActivE SOIL VENTILATION
5. HOLLOW BLOCK WALL VENTILATION
6. SUB-SLAB VENTILATION
7. WALL VENTILATION + SUB-SLAB VENTILATION
8- DRAIN TILE SUCTION
HOUSE PRESSUR1ZATION
9. AVOIDANCE OF DEPRESSURIZATION
10. PRESSURIZATION
AlR CLEANERS
11. PARTICULATE REMOVAL DEVICES
12. RADON GAS REMOVAL DEVICES
PASSIVE SOIL VENTILATION
13. SUB-SLAB VENTILATION
WELL WATER. TREATMENT
14. ACTIVATED CARBON SORPTION-, OTHER
- 8 -
-------
TABLE 1 (continued)
HOUSE DESIGN DETAILS
10.
ONE STORY vs* TWO STORY
BRICK 'VENEER VS- FRAME
FIREPLACE STRUCTURE vs- NO FIREPLACE
HALF-BASEMENT OR SLAB-ON-GRADE ADJOINING A FULL
ASEMENT/ VS. SUBSTRUCTURE ALL ONE LEVEL
LOCK VS- POURED CONCRETE FOUNDATION FOR SLAB ON
GRADE AND CRAWL SPACE HOUSES
WALL/SLAB OPENINGS ACCESSIBLE FOR SEALING vs*
NOT ACCESSIBLE
INTERIOR BLOCK WALLS IN BASEMENT vs. NO INTERIOR
BLOCK WALLS
EXTENT OF DRAIN TILE SYSTEM (COMPLETE vs. PARTIAL)
DRAIN TILE SYSTEM DESIGN (DRAIN TO ABOVE-GRADE
DISCHARGE/ VS. DRAIN TO INTERNAL SUMP)
FINISHED vs. UNFINISHED BASEMENT
NOTE: NOT ALL OF THESE ALTERNATIVES WOULD BE CONSIDERED
FOR ANY ONE MITIGATION/SUBSTRUCTURE COMBINATION; AT MOST,
THREE ARE CONSIDERED FOR A GIVEN COMBINATION, DEPENDING
UPON THEIR POTENTIAL IMPORTANCE ON MITIGATION DESIGN AND
PERFORMANCE.
IMTIAL RADON CONCENTRATIONS
3.
"LOW* - LESS THAN 0-1 WL
INTERMEDIATE - 0.1 - 1.0 WL
HIGH - ABOVE 1.0 WL
USUALLY NO MORE THAN TWO LEVELS OF RADON CONCENTRATION
CONSIDERED, FOR ANY ONE MITIGATION/SUBSTRUCTURE COMBINATION.
GE PL DG i CAL /METEOROLOGICAL Count TT
1- CONDITIONS RESULTING IN HIGH SOIL PERMEABILITY
2. CONDITIONS RESULTING IN LOW SOIL PERMEABILITY
3. POSSIBLY OTHER FACTORS
TECHNIQUE PERISH/OPERATING CONDITIONS
1. REDUCED SEALING IN CONJUNCTION WITH SOIL VENTILATION
TECHNIQUES/ VS. INCREASED SEALING
2* INDIVIDUAL POINT BLOCK WALL VENTILATION vs. BASEBOARD
APPROACH
3> SUB-SLAB SUCTION BY INDIVIDUAL POINTS vs. SUCTION
ON SUMP
4. ACTIVE SOIL VENTILATION OPERATED IN SUCTION vs* PRESSUKJ/.M
-------
TABLE 2
RADON MITIGATION TEST MATRIX
RANGE
INSTRUCTION)
S OF VARIABLES (NEW CONSTRUC
MlNG CONSI&tRED IN1TJALLY)
i-
HOUSE SUBSTRUCTURE TYPE
1. BASEMENT - HOLLOW BLOCK FOUNDATION WALLS
2» BASEMENT - POURED CONCRETE FOUNDATION WALLS
3. SLAB-ON-GRADE
4. CRAWL SPACE,
5. BASEMENT - POURED CONCRETE FOUNDATION WALLS/ WITH
ADJOINING HALF-BASEMENT OR SLAB'ON-QRADE
TECHNIQUE
1- FOR POURED CONCRETE FOUNDATIONS: THICK PLASTIC LINING
BETWEEN AGGREGATE AND CONCRETE SLAB; SLAB AND FOOTINGS
MONOLITHIC POURj UTILITY PENETRATIONS CAREFULLY SEALED;
ANY FIREPLACE STRUCTURE BUILT TO AVOID LEAKAGE, THERMAL
BYPASSING*
2. FOR HOLLOW BLOCK FOUNDATIONS: AS ABOVE FOR POURED CON-
CRETE FOUNDATIONS; SOLID CAP BLOCK AT TOP OF FOUNDATION
WALL; BLOCK BELOW GRADE COVERED WITH PLASTIC BARRIER ON
OUTSIDE' FACE, COATED TO REDUCE POROSITY ON INSIDE FACE;
GAP BETWEEN BLOCK AND BRICK VENEER MORTARED SHUT*
3- OTHER STEPS AS APPROPRIATE*
HOUSE DESIGN DETAIL
1. FIREPLACE STRUCTURE vs. NO FIREPLACE*
2» ENERGY EFFICIENT CONSTRUCTION vs. NORMAL
GEOL PIS i CAL/METEOROLOGICAL CONDJTIONS
I:
CONDITIONS RESULTING IN HIGH SOIL PERMEABILITY*
CONDITIONS RESULTING IN LOW SOIL PERMEABILITY*
- 10 -
-------
'changes also significantly reduce the number of "House Design Details"
which can affect soil gas entry and mitigation design/performance.
In reviewing Tables 1 and 2, it is clear that a lot of variables can be
important. It is apparent a priori that any attempt to address these
variables, even in a reduced manner, is likely to add up to a lot of houses
to be tested.
Matrix Format
The basic format of the matrix is presented in Table 3- Two of the six
categories of independent variables are shown explicitly in Table 3: House
Substructure Type defines the columns, and Mitigation Technique defines the
rows. (The number of variables used in this table for these two categories
reflect those identified in Table 1 for existing houses—5 substructure
ypes, 14 mitigation techniques.) For each cell within the matrix—i.e.,
for each combination of substructure type and mitigation technique—a number
of houses to be tested is identified (Nij, the. number of houses of sub-
structure type j to be tested using mitigation technique i). The whole
purpose of the matrix exercise, of course, is to derive reasonable values
of each Nij.
The value of Nij for a given cell will depend upon the extent to which
the other four categories of independent variables are addressed for that
particular substructure/mitigation technique combination. As discussed
previously, not all values of the other variables would be pertinent to a
given cell. For example, in no case would all 10 House Design Detail variables
for existing houses in Table 1 be applicable to a single substructure/mitigation
ination; in the estimates to date, a maximum number of 3 House Design Detail
ciriables have applied to any single cell.
- 11 -
-------
TABLE 3
RADON MITIGATION TEST MATRIX
MATRIX FORMAT
HOUSF SUSTRUCTURE TYPE
MITIGATION TECHNIQUE TYPE 1
TECHNIQUE 1 ; NU
TECHNIQUE 2 N21
YPE 2 ... TYPE.
N!2 N}5
N22
5
COMMENTS
TECHNIQUE
IS THE NUMBER OF HOUSES OF TYPE J TO BE TESTED USING
TECHNIQUE J
COMMENTS COLUMN WOULD PRESENT THE LEVELS OF THE DIFFERENT
INDEPENDENT VARIABLES USED IN DERIVING NJJ.
EACH NU WOULD INCLUDE SOME HOUSES WITH NORMAL DIAGNOSTIC
TESTING AND SOME WITH DETAILED DIAGNOSTICS.
MAXIMUM NUMBER OF INDEPENDENT VARIABLES USED IN CALCULATING
EACH fjjji
HOUSE DESIGN DETAIL ~ 10 VARIABLES, 2 LEVELS EACH
INITIAL RADON CONCENTRATION - 3 LEVELS
GEOLOGICAL/METEOROLOGICAL CONDITIONS - 2 LEVELS
MITIGATION DESIGN/OPERATING CONDITIONS - 4 VARIABLES, 2 LEVELS
EACH
- 12 -
-------
Each value of Nij would include some number of houses—perhaps 10 to 20
percent of the total—which would entail detailed diagnostic testing.
Example of Nij Calculation
Perhaps the most effective method for illustrating the approach to be
used in developing the matrix, is to show a sample derivation of one of the
values of Nij. Such a sample derivation is presented in Table 4.
For this example, the selected house substructure type is a basement
house with concrete 'block foundation walls (Type 1 from Table 1 for existing
houses). The selected mitigation technique is drain tile suction (Technique 8
from Table 1). Thus, for this example, Nij = N81
The first step in the derivation as shown in Table 4, is to determine
ich variables within the other four categories are potentially important
for this particular cell. Under the category of House Design Detail, three
variables of potential interest have been identified for this cell to date.
The completeness of the existing drain tile loop can clearly be important,
since'an incomplete loop might leave part of the foundation less well treated
by the suction. The, design of the drain tile system is important; if it
drains to an above-grade discharge away from the house, the design of the
mitigation system will be different from the case where the tiles drain to
a sump inside the house. The presence of an interior block wall (which
pentrates the slab and rests on footings) can be important; such interior
i
footings normally do not have drain tiles laid beside them, and the interior
wall (which can be ah important soil gas entry route) might not be adequately
treated unless the suction on the perimeter tiles extends effectively under-
iath the entire slab.'
- 13 -
-------
TABLE 4
. RADON MITIGATION TEST MATRIX
•APPROACH (EXAMPLE OF N, .Cfl) C.ULATION)
SELECTED ELEMENT FROM MATRIX
MITIGATION TECHNIQUE: 8 (DRAIN TILE SUCTION)
HOUSE SUBSTRUCTURE TYPE: 1 (CONCRETE. BLOCK BASEMENT)
STEP Ii IDENTIFY LEVELS OF EACH INDEPENDENT VARIABLE TO BE
CONSIDERED
HOUSE DESIGN DETAIL:
COMPLETENESS OF DRAIN TILE LOOP: 2 LEVELS
(COMPLETE, NOT COMPLETE)
DESIGN OF DRAIN TILE SYSTEM: 2 LEVELS
(DRAIN TO SOAK-AWAY, DRAIN TO SUMP)
INTERIOR BLOCK WALL IN BASEMENT: 2 LEVELS
(WALL, NO WALL)
INITIAL RADON CONCENTRATION: 2 LEVELS
(LOW, HIGH)
GEOLOGICAL/METEOROLOGICAL; 2 LEVELS
(LOW, HIGH SOIL PERMEABILITY)
filTIGATION'TECHNIOUE DESIGN: 2 LEVELS
(EXTENSIVE SEALING, LESS EXTENSIVE)
STEP 2: DERIVE FRACTIONAL FACTORIAL EXPERIMENT
FULL FACTORIAL EXPERIMENT WOULD BE 2X2X2X2X2X2=64 SETS OF
CONDITIONS
1/4 FRACTIONAL FACTORIAL - 1/4 X 64 - ]b SETS OF
CONDITIONS
A LARGER FRACTION (1/2) WOULD HAVE BEEN USED IF THERE HAD
BEEN ONLY 5 VARIABLES INSTEAD OF 6-
- 14 -
-------
TABLE 4 (continued)
STEP 3; Tg*T 5 HOUSES (I.E.. 5 REPLICATES) AT EACH SET tie
CONDITIONS
16 SETS OF CONDITIONS X 5 HOUSES/SET - 80 HOUSES
Nfcl - 80
FIVE REPLICATES WILL INDICATE THE VARIABILITY IN THE DATA.
(A FEW OF THESE HOUSES WOULD INCLUDE DETAILED DIAGNOSTIC
TESTING.)
STEP
EVALUATE THE DATA. DETRMINE FUER ACTION
IS THE CONFIDENCE INTERVAL NARROW ENOUGH SUCH THAT THE
AGENCY WOULD FEEL COMFORTABLE SUGGESTING THE TECHNIQUE
TO OTHER HOHEOWNERS HAVING SIMILAR CONSTRUCTION DETAIL/
RADON LEVEL/GEOLOGICAL CONDITIONS?
IF YES. NO FURTHER TESTING AT THE GIVEN SET OF
CONDITIONS IS NECESSARY
IF Hfb DETERMINE WHAT FURTHER TESTING IS WARRANTED
TO NARROW THE INTERVAL:
—' FURTHER REPLICATION AT THE GIVEN SET OF
CONDITIONS, TO BETTER DEFINE THE DISTRIBUTION
—: IDENTIFICATION OF OTHER VARIABLES WHICH
MIGHT BE RESPONSIBLE FOR THE DATA VARIABILITY,
INCORPORATION OF THOSE VARIABLES INTO THE
1 MATRIX.
- 15 -
-------
Two levels of initial radon concentration will be considered, low and high.
The low level is considered because drain tile suction is inexpensive and
i
i
aesthetic enough that it can reasonably be considered even when the radon levels
in the house are not significantly elevated. On the other hand, the high
level is considered because this technique is also effective enough that it '
can be considered evten for worst-case houses (if they have drain tile systems
•in place); Since both high and low levels are being considered, it is "not
felt necessary to address the intermediate level. In being applicable to
both high and low levels,. drain tile suction is unique; most techniques
capable of the high reductions needed for worst-case houses may be too
expensive to be considered by homeowners facing a low degree of mitigation
urgency. Thus, most techniques would be tested only at low and intermediate,
or only at intermediate and high, initial concentrations.
In the category|of Geological/Meteorological Conditions, soil permeability
is considered to be the variable of importance, with two levels (high and low)
considered. The success of drain tile suction depends upon its ability to
draw soil gas away from potential entry routes into the house, and the
permeability of the soil could potentially affect this ability.
In the category 'of mitigation technique design, one variable is considered
(the degree of sealing of openings between the house and the soil). . Such
sealing is a mitigation technique in itself which can often be important in
conjunction with other techniques. The naturally reduced pressures that
typically exist within houses serve as a "pump" sucking radon-containing soil
gas into the house, tending to work against the drain tile suction system.
i • -
To the extent that house/soil connections are sealed, 'this tendency
work against the drain tile suction system is reduced.
.- 16 -
-------
Reviewing the above paragraphs, it is seen that, at this time, six
variables of interest have been identified among the four other categories
of independent variables, for this case of drain tile suction in concrete
block basement houses. Two levels are being considered for each of the
variables.
The second step in the calculation of N81 is the derivation of the
fractional factorial experimental matrix for these variables. If a full
factorial experiment were to be conducted—considering every possible
combination of both levels of all six variables—then the number of sets
of conditions that would have to be tested would be 2 to the sixth power,
or 64. (One "set of conditions" would be, for example, a complete drain
tile loop draining to an external soak-away in a house having an interior
11 and a high initial radon level, on soil of low permeability, where
extensive sealing was used in conjunction with the suction system.) In
order to reduce the number of sets of conditions to be tested, statistics
would be utilized to design a fractional factorial matrix. The objective
of the fractional design is to enable separation of the effects of each
variable without having to perform the full factorial. The specific sets
of conditions which are selected for the fractional factorial are not
arbitrary, but must be picked with careful statistical consideration. The
compromise that one accepts when using a fractional design is that, as a
result of interactions between the variables, the separation of the
individual effects might not be possible with the same level of confidence
as a full factorial would provide.
For the purposes of estimating N81—with two levels of each of six variables-
is currently assumed that a one-quarter fractional factorial can be utilized.
- 17 -
-------
Thus, the number of sets of conditions to be tested would be
1/4 x 64 = 16 sets of conditions.
With some of the other Nij cells, the number of variables of interest will
be fewer than 6, and a 1/4 fractional factorial might therefore be too small to
provide sufficient'power to separate effects. In the current planning effort,
it is assumed that if there are 3 variables (2 levels each), a one-half
i
fractional factorial will be needed; if there are only two variables, a full
factorial would be performed.
After the number of sets of conditions is identified, the initial value
of H81 can be calculated by considering the number of houses to be tested at
each set of conditions. This number per set of conditions must be large
enough to give some reasonable measure of the data variability at that set
of conditions, since the underlying intent is to narrow the confidence interval
to some goal value. However, this number should not be too large, because we
wish to reach the goals with a minimum number of tests. For the purposes of
this planning, an initial number of 5 houses per set of conditions has been
selected. These 5 will provide an initial indication of variability/confidence
interval, so that a decision can then be made regarding what further testing,
if any, is warranted at that set of conditions. Thus, the initial value of
N81 is:
N81 = 5 x 16 = 80.
This number has been referred to here as the initial value of N81. The
ultimate value would be derived during the course of testing, as described
below.
- 18 -
-------
When the testing on five houses at a given set of conditions is completed,
the data would be evaluated statistically. These results might be pictured as
a plot of the fraction of houses tested versus the final radon concentration (or
the percent radon reduction), defining some type of distribution. If the
selected confidence interval for this distribution, apparent from these 5 houses,
is sufficiently, narrow that it falls within our goal value, then no further
testing at that.set of conditions is necessary. That is, the results from
those first 5 houses were consistent enough such that we feel confident that we
understand how that mitigation technique will perform in that house substructure
type at that set of conditions. It is expected that, in fact, some of the sets
of conditions will be satisfactorily addressed by the first 5 hosues.
With some other sets of conditions, however, the confidence interval ,
resulting from the first 5 houses will undoubtedly be too wide. In those cases,
it must be decided why it is too wide before deciding on the future course of
action. In some cases, the interval will be too wide simply because the 5
data points do not adequately define the distribution; the statistical formulae
calculate a~large confidence interval (a low degree of certainty) due to the
uncertainty in what the distribution really is. In such cases, it will sometimes
be appropriate simply to test some additional houses at the same set of conditions,
to better define the distribution. With the better-defined distribution, the
confidence interval might narrow to the extent desired.
i
In other cases, the breadth of the interval might be due to inherent
variability (which means, due to the presence of other variables which are not
explicitly addressed in the matrix but which have an important influence on
the observed results). In such cases, simply testing additional houses at
that set of conditions will do no good; even if an infinite number of houses
-------
were tested, and the distribution perfectly defined, the confidence interval
would still be too large. In those cases, the "hidden" variables must be
identified (by .rigorous inspection of the first 5 houses, if necessary), and
that cell of the matrix redesigned as necessary to incorporate the new
variables.
From the above discussion, it is apparent that, if anything, the initial
value for N81 will increase as the testing proceeds.
Status
The framework for the matrix has been defined, as described previously.
i
Detailed discussions between the engineering and statistical staff are underway
to more completely define the matrix, and to derive the values for each Nij.
In addition, a preliminary effort has been conducted to prioritize the matrix—
i.e., to suggest which cells, and which sets of conditions within each cell,
should be addressed first. This preliminary prioritization effort is described
in a latar section.
Very preliminary estimates have been made of the total numbers of houses
that might be needed to fill out the existing house and new house matrices
{i.e., the sum of all of the initial Nij's). Assuming 1/4 to 1/2 fractional
factorials and 5 replicates per set of conditions, as discussed in the previous
section, this preliminary total for initial coverage came out to be about 600
existing houses and about 100 new houses. About 15% of the houses would involve
detailed diagnostics. These numbers sound large, but considering the number
of variables involved, this size is not unreasonable from a technical stand-
point. J
- 20 -
-------
As further understanding is gained, it might be possible to intelligently
cut out certain variables or otherwise direct the program in a manner that will
reduce the number of houses required. In addition, in some cases, one house
can serve to address two or more data points on the matrix, at a reduction in
cost compared to two different houses; for example, referring to the earlier
example with drain tile suction, the conditions of extensive sealing and less
extensive sealing can be tested in a single house, with only the incremental
cost of additional sealing between conditions* However, on the other hand,
. the likelihood is significant that additional replications will be needed in
order to narrow confidence intervals in many cases, or that additional variables
will be identified. Thus, it seems reasonable to assume at this point that
any net change in the estimates of the number of houses will more likely be
in the direction of more houses.
Preliminary Effort to Prioritize Matrix Cells
A very preliminary effort is underway to obtain some prioritization of
house substructure types for the purposes of the matrix. The effort consists
of overlaying what we know about geographical substructure distribution, on
top of gross estimates of the distribution of radon-prone lands. The intent
is to identify whether particular substructure types appear to be more
prevalent in areas where the risk of elevated radon levels is relatively
greater. Such substructure types could then warrant higher priority in the
testing effort. Other information which can be obtained from this assessment
includes the geographical distribution of high-risk houses of a given sub-
structure type (suggesting possible sites that might be considered for con-
ducting the testing). •
•i
It is emphasized that this preliminary prioritization effort is intended
- 21 -
-------
to give very rough estimates, doing the best we can with what information is
available*
The approach being employed in this effort is described below.
!
First, an estimate is developed of the geographical distribution of
substructure types. The total number of housing units by state was obtained -
from the 1980 census; this will be updated to 1985 by considering building
permits issued after 1980. The breakdown of substructure types within each
state is being estimated using data obtained annually by the National Association
of Home Builders (NAHB), which gives this breakdown for the houses built in each
state after 1974. Multiplying the NAHB percentages times the census housing
unit totals yields an estimate of how many units of each substructure type
exist in a given state* There are a number of uncertainties built into this
estimation approach, among the key ones of which are: a) the uncertainty
regarding whether the NAHB data for a given year, which might be obtained from
i
only perhaps 20% of the houses built in a given state, in fact represents the
distribution among .all houses built in the state during that year; and b)
the uncertainty regarding whether the NAHB distribution for 1974-1985 in fact
»
represents the distribution among housing units built prior to 1974.
The second step in the approach is to estimate the geographical distribution
of high-risk lands. For the purpose of this effort, "high-risk lands" are
assumed to be those with both: a) an elevated level of uranium near the
surface of the ground; and b) a medium to high soil permeability, enabling
radon transport to the house. Data from the National Uranium Resource Evalua-
tion (NUBE) were used to estimate what percentage of the land area in each
state contained elevated near-surface soil uranium levels. A national map of
- 22 -
-------
surface geology type was used to estimate soil permeability, by assigning a
high, medium or low permeability to the various geology types; the percentage
of the land area in each state having high, medium and low permeability was
calculated. The percentage of each state having elevated potential was then
obtained by multiplying the MURE elevated uranium percentage times the high
permeability percentage (or the high plus medium permeability percentage).
The uncertainties in this approach for estimating radon risk are legion:
high-radon areas may exist where there is not elevated uranium within one
foot of the surface; the NURE data actually cover only a small percentage of
the nation's land area; high-uranium and high-permeability areas may not
randomly overlap, so that simple multiplication of those two percentages
might not give an accurate picture; and others. However, this approach is
sed as a first approximation.
The last step* in the approach is to calculate how many high-risk houses
of each substructure type are in each state. This calculation is made by
multiplying the percentage of high-risk land in each state times the total
number of houses of each substructure in that state. This approximation
assumes that the houses of all substructure types are uniformly distributed
over all of the land area in the state.
\This analysis is not yet complete. However, some initial results are
presented in Table 5. As an example of how to read this table, the top
entry indicates that Alabama has about 10% of its land area containing
elevated near-surface uranium deposits, and about 25% having geologies
considered highly permeable. Thus, the percentage of Alabama land assumed to
have the potential for elevated radon is 10% x 25% = 2.5%. Since Alabama
had 1,073,053 housing units in the 1980 census—of which 10.5% were basement,
- 23 -
-------
TABLE 5
Gross estimate of houses, by substructure type, In areas with risk of
elevated radon levels
PERCiNf OF
LAUD W/
•COTERM- URANIUM
INDUS DEPOSITS
STATES (NURf MAP)
AL
AZ
AR
CA
CO
CT
OE
rt
CA
ID
11
IN
IA
KS
ICT
LA
ME
MJ>
MA
Ml
MM
MS
HO
MT
NE
NV
NH
WJ
NH
NT
NC
NO
OH
OK
Oft
PA
R!
SC
.SB
TN
TX
UT
VT
VA
WA
•wv
Wt
WY
TOTAL
X Of TOTAL
10*
6X
n
4X
10X
zox
SX
, 33X
11X
5X
OX
10X
Ox
1SX
8X
4X
sx
10X
8X
n
ox
ox
10X
' 10X
: 10X
sx
3X
ZOX
5X
SX
10X
sx
ox
zox
2X
; sx
> 20X
ZOX
15X
35X
n
: isx
1QX
10X
10X
ox
sx
30X
PERMEABILITY (PERM)*
LOU
ISX
ox
20X
m
ox
ox
ox
ox
tox
ox
ox
10X
ox
10X
tOX
ox
ox
tOX
ox
ox
ox
sx
36X
ox
ox
ox
ox
10X
sx
ox
4 OX
20X
ox
(OX
10X
zox
ox
SOX
sx
2SX
15X
ox
ox
SOX
1SX
ox
ox
ox
HEP HIGH
60X
tSX
60X
t5X
45X
100X
100X
3»X
tox
35X
BOX
cox
9ox
rox
SOX
80X
90X
tOX
10QX
90X
70X
95X
47X
m
rox
35X
100X
60X
3SX
asx
zox
60X
-60X
SOX
SOX
30X
100X
2ox
MX
40X
in
tox
100X
10X
6SX
ox
TOX
35X
Z5X
S5X
ZOX
40X
ssx
ox
ox
6JX
zox
65X
ZOX
tox
tox
zox
SOX
zox
10X
ZOX
ox
10X
sox
ox
17X
23X
30X
6SX
ox
SOX
MX
1SX
tox
zox
tox
SOX
tox
SOX
ox
SOX
sx
ssx
sex
60X
ox
tox
zox
100X
SOX
63X
POTENTIAL
POTENTIAL ELEVATED POTENTIAL
POTENTIAL UEVATEO *ADON ELEVATED
ELEVATED RADON LEVEL RAOON
RADON LEVEL CRAUL LEVEL
LCVEL 8A$£MfNT JPACI SLAB
HOMES HOMES HOMES HOMES
(NURE) (NURE MAP) (HUftE MAP) (NURE MAP) (NURE MAP)
Z.5X
3.3X
1.4X
1.6X
5.5X
o.ox
o.ox
Z1.SX
z.zx
3.3X
O.OX
1.0X
o.ox
3. OX
2.4X
0.6X
o.sx
2. OX
O.OX
o.sx
o.ox
D.OX
i.rx
Z.3X
3. OX
3.3X
O.OX
6.0X
3. OX
0.6X
t.OX
1.0X
O.OX
«.ox
0.8X
t.OX
o.ox
10. OX
o.sx
17.3X
2.7X
9. OX
O.OX
t.OX
2. OX
O.OX
1.5X
19.SX
•
26
21
*
Si
39
SOS
30
8
IS
Zl
23
8
1
15
7
23
t
13
S
as
9
19
64
1
56
S
96
81
1
153
• 98
Z9
52
22
17
21
1,684
.826
.133
,601
,1Z2
,951
0
0
,«2
,005
,'.77
0
,069
0
,351
,533
,tZ7
,579
.370
0
.397
0
0
,647
.693
,758
.507
0
.MS
,586
,501
,691
,617
0
,572
.695
.296
0
,*65
,tlf
,650
,76t
.625
0
.368
,Z51
0
,837
,293
,293
3.2X
2,817
211
91Z
9,67t
31,961
0
0
0
12,002
4,743
0
6, tot
0
20,t97
I6,9tt
29S
1,282
1t,UO
0
7.138
0
0
22.no
4,622
1Z.8M
28
0
44,650
Ut
15.894
10.674
1,537
0
1,697
1,310
88.592
0
6,517
1,286
t3,770
1.481
28,143
0
24,613
13.907
0
16,U3
19.2H)
t88,291
29.0X
8.
3,
16,
7.
25,
6,
3,
3,
t,
2.
55.
36,
«,
3,
50.
67.
13.
7.
1.
307,
182
317
937
404
591
0
0
292
751
312
0
466
0
641
354
42
28
384
0
222
0
0
473
70
688
533
0
204
2M
760
227
to
0
566
129
832
0
$08
30
606
0
889
0
092
788
0
171
916
799
18.3X
1S.8ZB
20,605
4,753
58,044
. 400
D
0
480.541
11,252
123
0
5.199
0
214
2,236
8,090
69
845
0
37
0
0
1,064
0
206
2,946
0
6.011
9,155
2,828
17,790
to
0
54.310
256
3,852
0
24,439
101
t2.25t
97.282
S9Z
0
14,663
556
0
1,516
106 •
888,203
32. 7X
- 24 -
-------
30.5% were crawl: space and 59% were slab—multiplying 2.5% times the number
of units with each of these substructures yields the estimated numbers of
houses of each type having the potential for elevated levels under the
assumptions used here (e.g., 2,817 basement houses).
It is noted in Table 5 that—from the national totals—this analysis
suggests that slab-on-grade houses and basement houses have the greatest
!
number of units in potentially radon-prone areas. The high representation by
slab houses results from the large contribution from Florida, which has a
relatively large percentage of high-radon-potential land area and involves
slab construction almost exclusively. The lower number of radon-prone crawl
space units is not surprising, since the total number of crawl space units
nationwide is limited (15% of the total). It is emphasized that this analysis
is attempting to develop gross estimates of the number of each house type
built in areas with elevated risks for indoor radon; it cannot at this time
predict what the ^distribution of radon levels inside the houses might be.
For example, it cannot account for the impact of substructure type of actual
indoor levels; due to the different degrees of house/soil contact, a basement
house might often be expected to experience higher indoor levels in a given
location than would a crawl space house.
The results of this analysis generally support EPA's current emphasis
on basement and split-level (basement plus slab) houses, and suggest that
further attention to slab houses might be in order. However, this result
is one which might have been'expected a priori, since the greatest number of
houses nationwide are basement (50% of all units) and slab (35%). For this
analysis to be more useful in directing the future program, the next steps
in completing the analysis will address a finer breakdown of substructure types
- 25 -
-------
e.g., block foundation wall basements and poured concrete foundation wall
basements), and possibly combinations of substructure types (e.g., basement
plus slab-on-grade or split levels). In the longer term, it would be desirable
to attempt to estimate the distribution of actual indoor radon levels as a
function of substructure type, if a meaningful method for making this estimation
could be identified.
Efforts are continuing now to complete this preliminary study. In
addition to expanding the number of substructure types addressed, as discussed
above, the on-going work includes: refined estimation of the current number
of units nationwide for each of the substructure types (by updating the census
figures and more completely drawing upon the NAHB data base); and overlaying
of the substructure type and the elevated-radon-potential land information on
i
geographical unit finer than the state level (e.g., by county or zip code),
in an effort to more accurately match the two.
Technical Issues Needing Review
There are a number of technical issues which need to be reviewed by the
Mitigation Subcommittee of the Radiation Advisory Committee of the Science
Advisory Board. Some of these are listed in this section. The Subcommittee
will undoubtedly have additional questions they will want to raise (and
answer). The matrix is evolving and is still under intensive review by the
Air and Energy Engineering Research Laboratory. Suggestions by the Subcommittee
will be incorporated in the next iteration to be prepared as soon as the
Subcommittee recommendations are received.
t. Does the basic approach for the development of the matrix appear
reasonable (fractional factorial design, 5 initial replicates per
condition, further testing determined from the initial 5 tests)?
26
-------
2. Do the selected independent variables (Tables 1 and 2) appear reasonable?
3. How narrow should the confidence interval be before testing is
stopped? What confidence interval should be used (e.g., 68%, 95%}?
4. Regarding efforts to prioritize the matrix:
a. Should initial focus be on conditions resulting in most acute
exposure, or greatest cumulative exposure, or both?
b. What other approaches for prioritization, in addition to the
initial effort described in the previous section, might be
considered?
c. How might estimates be derived indicating the distribution of
indoor radon concentrations for each substructure type?
- 27 -
------- |