United Statoa
Environmantal Protsction
Ag«ncy
QWiceof
Toxic Substances
Washington, D.C. 20460
EPA 560/5-88-004
June 1988
Toxic Substance*
In  a
To
                   in Indoor  Air

-------
Bg/7'iq.' i .... . .. •. 	 .-..---. 	 	 i
REPORT DOCUMtNTATION »• •««*«• "^ gg0/5_88-004 *" V
., nil* »nd suu«iti» , , . ,
Preliminary Experiments 1n a Research House to investigate
Contaminant Migration in Indoor Air
''Author.,t Dcto
June 1988
a.
a. p
10.
n.
(C5
(0)
IS.
•wfonnina Onjanliatlon n«p(. No.
IE-1882
Pn>i»ct/Tatk/Work Unit Ho.
ControettC} or Qr«nt(O) N?.
68-02-4254
Type af Rocvort & f>*riod Cov«r«d
Final Report
14.
     The  EPA Project Officer was Elizabeth  Bryan;  the EPA Task Manager was Patrick
     Kennedy.            _^^		-
.«. Abstract (LImH: 200
     Controlled experiments were performed  in  an  unoccupied research house to provide
     m  a  detailed characterization of  the migration patterns of contaminants  released
     indoors from consumer products and  (2) a  basis for assessment of the exposure
     Implications of contaminant migration  and the accuracy of currently -used exposure
     assessment models.  To enable relatively  detailed spatial and temporal monitoring
     with readily available instrumentation, carbon monoxide (CO) was chosen as a
     surroqate contaminant for the investigation.  A point source was simulated In  the
     master bedroom by releasing CO from a  pressurized tank through a pnematic  line over
     a  1. 25-hour period.

     During the release period, measured CO concentrations typically were  3  to  4 times
     hiaher In the release area than  in  other  areas on the floor of  release.- Within an
     hour after the release was terminated, concentrations approached spatial uniformity
     even though a central air circulation  fan that would have promoted  contaminant
     migration was turned off as part of the experimental design.  A single-chamber
      Indoor air quality model provided closer approximation of passive  than  active
      exposures.  Use of a two-chamber model resulted  in better estimates of  each type of
      exposure. *                             ,
17. Doeymant Analytl* », D«eriptefll

      Exposure Assessment
      Indoor Air Quality Monitoring
      Indoor Air Quality Modeling
                  T«rm»
      Contaminant  Migration Indoors
      Active Exposure
      Passive  Exposure
   e. CO1ATI Flald/Qraup
IS. Availability Stateman;
       Distribution Unlimited
19. S*eurft» Claw CTH*» *wort}
    Unclassified
                                                      ZO. S«curtty Ctosa ff»a« •"*«»>
                                                          Unclassifie
21. No. o) P*8
       85
                                                                              Z2.
S«< AN31-Z39.1S)
                                       Sea
                        (MTtonAt. romi 27:
                        (Fe*Tt«riy HTIS-35J

-------
                                     EPA 560/5-88-004
                                     JUNE 1988
         PRELIMINARY EXPERIMENTS
           IN A RESEARCH HOUSE
        TO  INVESTIGATE CONTAMINANT
         MIGRATION  IN INDOOR AIR
                    by

    Michael D.  Koontz,  Harry  E.  Rector,
     Roy C.  Fortmann, Nlren L. Nagda
       EPA Contract No.  68-02-4254
             Project Officer

            Elizabeth F.  Bryan
       Exposure Evaluation Division
        Office of Toxic Substances
         Washington, D.C,  20460
   U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF PESTICIDES AND TOXIC SUBSTANCES
         WASHINGTON, D.C.  20460

-------
                            DISCLAIMER
    This document has been reviewed and approved for publication
by the Office of Toxic Substances, Office of Pesticides and Toxic
Substances, U.S. Environmental Protection Agency.  The use of
trade names or commercial products does not constitute Agency
endorsement or recommendation for use.
                                 ii

-------
                         ACKNOWLEDGMENTS
    This report r=prsp=rea by GEOMET Technologies ,         _
Germantown, Wara"dl,!?L^e ^posure Assessment Branch (EAB),
                      1
    in addition to the authors of this report  a

                                             11
                 i.
Program Management -

Task Management -


                4.
Technical Support -



Editing -
                    i
Secretarial/Clerical
                                       Gavaneh  Contos,  Versar
                                       oaycui*

                                       H.  Lee Schultz,  Versar
                                       Nlren Nagda,  GEOMET

                                       Donald Cade,  GEOMET
                                       David Skidmore,  GEOMET
                                       Laura Mehegan, GEOMET

                                       jo Ann Koffman,  GEOMET

                                        Jean Fvock, GEOMET
                                        jeanette Behnke, GEOMET
                                        Dana Cue, GEOMET
                                  ill

-------
                        TABLE OF CONTENTS



                                                           Page

    Executive Summary                                       ^x

1.   Introduction                                             *

l.l.     Background                                          1
1.2.     Objectives and Scope                                2

2.   Experimental Design and Research Methods                 5

2,1,     Research Setting                                    5
2.2.     Contaminant Release and Monitoring                  5
2.3.     Ancillary Measurements                             14
2,4.     Quality Assurance and Control Procedures           14
2.5.     Data Processing and Analysis Procedures            20

3,   Analysis of Experimental Results                        23

3.1.     Experimental Conditions                            23
3.2.     Data Quality                                       27
3.3,     Concentration Profiles                             33
3.4.     Integration Across Experiments                     42

4.   Analytical Modeling                                     49

4.1.     Single-Chamber Mass Balance Model                  49
4.2.     Multiple-Chamber Modeling                          56

5.   Discussion                                              63

5.1.     General Perspective and Needs                      63
5.2.     Insights from the Current Investigation            67

6.   Conclusions and Recommendations                         73

6.1.     Conclusions                                        ?3
6.2.     Recommendations                                    74

7,   References                 .                             77

Appendix A.   Indoor Air Quality Modeling Concepts
              and Formulations                               ?9
          Preceding page blank

-------
                         LIST OF FIGURES                      Page


 1  Floor plan of GEOMET research houses.                       6

 2  Three possible stages of contaminant history.               7

 3  Spacing of protable continuous monitors to form a
    vertical sampling string.                                 10

 4  General locations of the release point, the vertical
    sampling planes,  and the stationary monitoring network,    12

 5  Daily checklist form.                                     18

 6  Operational checklist for experiments.                     19

 7  Functional relationships among vertical sampling strings
    of the detailed monitoring network and  stationary
    monitoring sites.                                         22

 8  Response of one CO detector to 0 and 9.06 ppm during
    multipoint calibrations and daily zero  and span checks.    32

 9  vertical concentration gradients with CO detectors
    arrayed in the master bedroom.                            34

10  Horizontal concentration gradients with CO detectors
    arrayed in the master bedroom.                            36

11  Vertical concentration gradients with CO detectors
    arrayed in adjacent bedrooms.                             37

12  Horizontal concentration gradients with CO detectors
    arrayed in adjacent bedrooms.                             39

13  vertical concentration gradients with CO detectors
    arrayed in the hallway.                                   40

14  Horizontal concentration gradients with CO detectors
    arrayed in the hallway,                                   41

15  Concentration profiles across all experiments at the
    anchor point.                                             44

16  Concentration profiles for each experiment at the
    midlevel anchor point in the hallway and stationary
    monitoring sites.                                         45
                               vil

         Preceding page blanfc

-------
                                                              Page
17  Spatial profile of CO concentrations upstairs during the  ^
    release period.
18  Spatial profile of CO concentrations upstairs following
    the release period.
19  Single-chamber mass balance model calculations for
    generalized experimental conditions.                      50
20  Comparison between single-zone model predictions and
    measurements near the release area,                       bl
21  Comparison between single-zone model predictions and
    measurements in the hallway and front bedrooms.           5J
22  comparison between single-zone model predictions and
    measurements at stationary monitoring locations that
    represent likely sites of passive exposures.              b4
23  Comparison between single-chamber model  predictions and
    volume-weighted average  indoor  concentrations  for three
    morning experiments.
24  Airflows used  as  inputs  to  a two-chamber model.            58
25  Comparison between two-chamber  model predictions and
       *"    -        i  _i_*__^	  j —  .*. A .mtt f-f ***r\ r* •£>*"*iv m/*"* V n "1 fl rf
    S_.i*JlllL/Q L, J.O Wii iXii*  V»£* wr " '>-' •»«**•«-...--—— _..- —   ^
    measured concentrations in each zone for morning
    experiment type 2.
 28   Proposed  array of PFT sources for future contaminant
     migration experiments.
                                                               59
 26  Overview  of consumer exposure model,                       64
    Illustrative monitorirv
    migration experiments.
27  illustrative monitoring array for  future  contaminant
                                                               69
                                viii

-------
                          LIST OF TABLES                      Page

 l  Characteristics of Release Scenario                        9
 2  Sequence of Contaminant-Release Experiments               13
 3  Specifications for Ancillary Measurement Parameters       15
 4  Instrumentation for Ancillary Measurement Parameters      16
 5  Summary of Prevailing Conditions During Each Experiment   24
 6  Summary of External Audit Results for Five CO. Analyzers   28
 7  Accuracy and Precision of Five CO Analyzers at Audit
    Concentration of 10.34 ppm                                28
 8  Accuracy, and Precision of all CO Analyzers Used in the
    Investigation at Final Calibration Input of 9.06 ppm      29
 9  Changes in Slopes and Intercepts Between Beginning and
    Ending Calibrations for CO Detectors Used in the
    Investigation                                             31
i.0  Comparisons of Model Estimates and Measured value? for
    Peak and Time-Weighted Average Concentration During and
    After Contaminant Release                                 60
11  Important Inputs to Model Lookups and Calculations and
    Associated Data Sources                                   65
12  Utility of PFT Data Base and Research House Experiments   66
                                ix

-------
                        EXECUTIVE SUMMARY


    Over recent years, advances have been made in the development
and application of methods for the assessment of exposure to
chemical contaminants released from consumer products to the
indoor environment.  In spite of these advances, there still exist
siqnificant gaps in our understanding of the behavior of
contaminants following their release and the implication of this
behavior for human exposure.  In particular, the extent of passive
exposure (i.e., that arising from contaminant migration to indoor
air spaces from the space where a contaminant is released) is
poorly understood.

    The OTS exposure-assessment process has recognized passive
exposure as an issue of concern; however, appropriate information
to provide quantitative treatment of this issue is not currently
available.  As a result, significant uncertainties exist with
reqard to the accuracy of Indoor air exposure estimates.  The
objectives of the investigation described in this report were {!)
to perform a detailed characterization of contaminant migration
patterns in an unoccupied research house through a series of
controlled experiments, (2) to assess the exposure implications of
contaminant migration, and  (3) to assess the accuracy of c^rently
used exposure assessment models and explore model refinements that
could lead, to improved estimates of active  and  passive  exposures.

    To enable detailed spatial and temporal monitoring  with
readily available  instrumentation, carbon monoxide (CO) was chosen
as a surrogate contaminant  for the investigation.  CO was released
from a point source  in the  master bedroom of the  research house at
a constant, known  rate over a period of  1.25 hours.  A  network  of
nine portable continuous  CO detectors was arrayed to measure
horizontal and vertical concentration gradients in each of  three
configurations—(1)  in the  release area,  (2)  down a  connecting
hallway,  and  (3)  in  entrances  to nearby  bedrooms.  An anchor
siring of three  vertically  arrayed detectors  located just outside
the entrance to  the  master  bedroom provided continuity  across the
three different  arrays of detectors.  In addition, a stationary
network of sampling  locations  representing  likely passive exposure
sites was  sampled  on a  rotating basis with  a nondlspersive
infrared  analyzer.   The  simulated  contaminant release was
performed one! in the morning and  once  in the afternoon for each
array of  CO detectors.   Ancillary  parameters such as outdoor CO
concentrations,  meteorological conditions,  and air  infiltration
rates were also monitored.
                                 xl
         Preceding page blank

-------
    During the release period,  measured  CO concentrations
typically were 3 to 4 times higher in the release  area than  in
other upstairs areas of the house.  However,  within 45 to  60
minutes after the release was terminated, concentrations
throughout the upstairs of the house approached spatial
uniformity, eventhough a central air circulation  fan that would
have promoted contaminant migration was  turned off as part of the
experimental design.  Some evidence of contaminant migration to
the downstairs living area of the house was observed during
forces is involved in the mixing and transport of contaminants.

    interestingly, a single-chamber model— similar to that
currently used in OTS assessments of active exposures in
residential environments-provided a closer approximation of
oassive than active exposures.  Use of a two-chamber model
?esSlted if better estimates of each type of exposure.  Thus  even
t hnnoh a comolex set of forces may underlie contaminant mixing and
transport patterns  tSS concept of treating general interchamber
lirrlnw natterns as a steady-state condition into which
consimer?pro^SI emissions are injected and transported appears
valid and useful for improving exposure estimates.

    To provide a basis for continued refinements and  improvement
to  currently used models  for  exposure  assessment s   aad itional
research hoSse experiments need to be  performed for a Ji*«

                                                      k f  f%
IxSeriSlntal  results can  be  easily compared.   Future  contaminant-
mifratTof SpeltSnts  should  include measurement of time-varying
and integrated  airflows  as  a  routine component.  In ^dition,  the
transferability  of  results  from  research  houses to different
hSSSin|types  should  be  assessed  by  replicating  selected  experi-
ments in a  limited  number of local  residences.

     To increase the applicability of exposure assessment  models to
          f,^
 "Saged airflow rates among selected zones of a residence should
 be analyzed as soon as these results are assembled in a
 computer-accessible format,
                                  xii

-------
    in parallel with expanded data collection, assimilation, and
analysis efforts, activities'to refine and improve current
exposure assessment models should be initiated.  This process
should begin with the development of a generalized multichamber
model and continue with refinements and expansions as critical
inputs are obtained through supplemental efforts such as those
recommended in this report.
                                 xiii

-------
           PRELIMINARY EXPERIMENTS IN A RESEARCH HOUSE
        TO INVESTIGATE CONTAMINANT MIGRATION IN INDOOR AIR
1,       INTRODUCTION


1.1.     Background

    Over recent years, advances have been made in the development
and application of methods for the assessment of exposure to
chemical contaminants released from consumer products to the
indoor environment.  An overall analytical structure has evolved
to guide such evaluations, as have a variety of methods for
calculating the indoor air contaminant concentrations to which
receptors are exposed.  Data bases quantifying consumer product
use patterns and the chemical makeup of many consumer products
have also been developed to support the application of the
assessment process.

    In spite of these steps, significant gaps exist in our
understanding of contaminant behavior following release and the
implication of this behavior for human exposure.  In particular,
the extent of passive exposure (i.e., that arising from
contaminant migration to indoor air spaces from the space where a
contaminant is released) is poorly understood.

    The OTS exposure assessment process has recognized passive
exposure as an issue of concern; however, appropriate information
to provide quantitative treatment of this issue is currently
lacking.  As a result, significant uncertainties exist regarding
the accuracy of indoor air exposure estimates.  Thus, specific
needs exist (1) to investigate the issue of passive exposure to
determine whether it warrants further attention and (2) if so, to
develop strategies for obtaining appropriate levels of
quantitation.

    In a recent report (GEOMST 1987a), average Interior airflows
and air infiltration rates measured with multiple perflourocarbon
tracers (PFTs) for a typical residence were used as inputs to a
multicharnber indoor air quality model.  This model was used to
demonstrate the implications of passive exposure to chemical
substances released from a consumer product within any of three
zones in the house.  For one of these cases, it was shown that
within 2 hours after the 10-minute release period, concentrations
were higher in another zone than in the zone where the substance
was released.  This analytical exercise demonstrated that the
issue of passive exposures to chemical substances released from
consumer products warrants concern and further investigation.

-------
1,2.     Objectives and Scope

    The results of the analysis described above indicated that a
data base maintained by Brookhaven National Laboratory,  which
contains results from PFT measurements in approximately  4,000
U.S. residences, may be a valuable input to future exposure
assessments concerning the use of consumer products in residential
environments.  However, the PFT measurement technique is generally
limited to quantifying average airflows over time periods of
several days or longer.  Consequently, the use of such data may
still introduce inaccuracies in exposure assessments because many
consumer products are used for shorter durations on the order of
hours or minutes.

    In concept, general airflow patterns derived from PFT
measurements can be treated as a steady-state or slowly changing
condition into which emissions from consumer products are injected
and transported.  Although this concept provides for facile
incorporation of readily available data, it represents an untested
extrapolation.

    Consequently/ a limited series of experiments was designed to
quantify contaminant migration over detailed temporal and spatial
scales in an unoccupied research house maintained by GEOMET.
These experiments were exploratory in nature and were Intended to
improve our understanding of underlying physical processes rather
than to mimic any specific exposure scenario.  The major
objectives of this effort were as follows:


         «    To examine the basic time scales and variations  for
              contaminant migration to adjacent airspaces;

         •    To examine the relative levels and durations of
              passive  exposure In adjacent airspaces due to
              contaminant migration;

              To examine the dilution effects of contaminant
              migration to adjacent rooms  and air exchange
              with the outdoors; and

         »    To assess the ability of single- or multi-chamber
              models to predict active and passive exposures.


    The experimental design involves  the controlled  release  of a
surrogate contaminant  in one room of  the research house  and
monitoring of  contaminant mixing and migration to other  rooms
through a detailed network of  sensors.  The  results  of these

-------
experiments will enable a quantitative assessment of the utility
of existing and readily obtainable data relating to interroom
airflows,

    The research setting, experimental design, and measurement
methods are described in Section 2 of this report.  Experimental
results are presented in Section 3 and modeled in Section 4.  The
implications of the results are discussed in Section 5 in terms of
model accuracies, applicability of existing data on interior
airflows,  and future research needs.  Conclusions and
recommendations stemming from this investigation are outlined in
Section 6.

-------
2.
EXPERIMENTAL DESIGN AND RESEARCH METHODS
    This section describes the setting in which experiments were
conducted, together with contaminant release and monitoring
methods.  Measurement techniques for ancillary parameters such as
outdoor meteorological conditions, indoor temperatures,  and air
infiltration rates are also described.  The section concludes with
a description of quality assurance/control, data processing, and
analysis procedures.  Portions of this section are extracted from
a recent GEOMET report (1987b) describing sampling and analytical
protocols for the investigation.


2.1.     Research Setting

    GEOMET's research house facility consists of two bilevel,
wood-frame houses that were constructed in the fall of 1982,  The
houses, located on adjacent lots in Gaithersburg, Maryland, are
identically oriented, facing 19° east of north  (i.e., north-
northeast).  Floor plans for the houses are shown in Figure 1.
The main living area is upstairs; the downstairs area is divided
into an unfinished living area and an Integral garage.  The total
upstairs living area of each house is 130 rn2  (1400 ft^).  The
upper and lower levels are connected by a  stairway with one
landing at the house entryway.

    The research houses were constructed using  "closed-wall"
techniques.  Siding, sheathing, insulation, vapor barrier,  and
windows were assembled at the factory to  form complete wall
panels.  The completed wall panels were  installed at  the building
site to form the building shell.  Insulation  for the  re-^arch
houses  features a continuous polyethylene  vapor barrier  with  glass
fiber batts between the wall joints  (R-value  of 11).  The  attic
contains  8 inches of loose fill insulation between the ceiling
joists  {R-value of  30).

    Abbreviations that are used later in this report  for selected
rooms or  areas of the house are also indicated  in Figure 1.


2.2.      Contaminant Release  and  Monitoring

    The experimental  strategy was designed to monitor contaminant
history from a constant  point source through  three possible stages
 (Figure 2};


          Source  cloud—in the immediate vicinity  of  the
          source,  concentrations are controlled by  the
          Preceding page Wank

-------
              -e&-
           Haittr Btdroon
              (HBR)
                 _">
^x!
                           T-T
                                    K lichen
                                            Ofnlns Room
                  /
           Cornir
           Bedroom
                    \
           -O=t-
  W7

   Front
  Bcdrooa
                                 Entry
                    Llvlnj Room
                                             UPST4IRS
Csrsac

^ Unftflijhtd
Living Ana


L
DOWNSTAIRS
0 5m
icale: 1 	 1
Figure  1.   Floor plan of GEOMET  research houses

-------
Source Cloud
 Transition
 Steady-State
Figure 2.  Three possible stages of contaminant history.

-------
         amount of material released and the dimensions of the
         source cloud;

     •    Transition—once the source cloud approaches the
         dimensions of the room, concentrations begin to
         be affected by air exchange and migration to
         adjacent airspaces; and

     •    Steady state—concentrations throughout the general
         airspace change in response to air exchange, room-
         to-room flows, and (as applicable) continuing
         emissions and chemical decay.


    The basic experimental sequence began with a controlled
release of a surrogate contaminant from the center of the master
bedroom in one of the research houses.  The measurement
strategy involved a set of portable continuous analyzers that were
arranged to form a sampling plane that was "walked" through the
concentration patterns created by controlled repetition of a
single-release scenario.  General characteristics of the release
scenario are summarized in Table 1.

    Carbon monoxide (CO) was selected as the surrogate
contaminant.  CO was released from a pressurized tank located
outdoors that contained approximately 1 percent CO in air.
Pneumatic lines were used to direct the contaminant from the tank
to the release point in the master bedroom; the gas feed was
controlled externally so that the technician would not need to
enter the house during the conduct of any experiment.  To prevent
undue mixing from momentum of the release flow, a,ceramic frit was
installed on the outlet at the release point.  Based on a
preliminary experiment, a release rate of 1.2 L/min and duration
of 1.25 hours 'were chosen; these conditions resulted in short-term
CO peaks on the order of 10 parts per million (ppm) in the master
bedroom and concentrations,below 5 ppm elsewhere in the house.

    Evolution of the source-cloud and transition stages was
monitored by arranging nine portable continuous CO monitors
(General Electric Model 15ECS3C03) to form a vertical sampling
plane.  The vertical sampling plane was made up of three vertical
sampling strings.  As illustrated in Figure 3, the nominal
floor-to-celling dimension of the upstairs of the research house
is 2.3 meters.  Sampling heights of 0.38, 1.16, and 1.93 meters
were specified for each vertical string to systematically divide
the vertical sampling plane into three layers that were each 0.77
meters deep.

-------
          Table l.   characteristics of Release Scenario


l.   General conditions

         All exterior openings (doors, windows) closed
         All interior doorways on main floor open
         Doorway to downstairs opened
         Doorway to garage closed
         Operation of central circulation fan suppressed

2.   Release conditions

    •    Location—geometric center of master bedroom
         Rate—1.2 L/min from tank containing 0.9844 percent CO
         in air
         Duration—1.25 hours

-------
                                                        Height
                                                       (meters)


Ceiling       .	.	—        2'31


Top level                     +                          1'93
                                                         1.54
Mid level                     +                          1<16

                  	   	i	,	,	        0.77



Low level                     +                          0<38


Floor	,	—	
Figure  3.  Spacing of portable  continuous monitors  to  form a
vertical sampling string.  Crosses  (+) denote monitoring  height
for top, raid,  and low levels  situated between floor and ceiling,
                                 10

-------
    To provide continuity from experiment to experiment,  °ne
locatioS was designated as the anchor point for all ^angements
of the vertical sampling strings.  This point (designated as point
A in Figure 4} was centrally located in the hallway equidistant
from the doorways of all three bedrooms .   For the initial
experiment  vertical strings 2 and 3 were set up to bracket tne
rSwsS^oint, giving a sailing plane defined by the anchor plus
positions S2 and S3 indicated in Figure 4.

    For experiment type 2, vertical strings 2 and 3 were relocated
to the hallway (positions H2 and H3 in Figure 4).  For experiment
type 3, the Sampling plane was folded to extend into the corner
bedroom and the front bedroom by placing sampling strings 2 and 3
at locations D2 and D3, respectively.

    Data from the nine portable monitors were collected  every
6 seconds and processed by a data logger for storage as  l -minute
averages during each experiment.

    A stationary sampling network was  operated  to  continually
measure CO concentrations in each major  room of the research
hoSse  except for the  room of release.   This second network, also
indicated 11 Figure 4, consisted of  a  single continuous  analyzer
(Beckman Model 886) that was automatically sequenced among  five
sampling points on a 3-minute schedule,  providing  a measurement  of
eS point every  15 minutes.  For consistency  wjjh previous
monitoring protocols,  indoor CO  measurements with  the  stationary
network were  taken at  a  height of  1.07 meters.

    Experiments were conducted during  the period December  1-4,
1987.  As  summarized in  Table  2, the controlled-release  scenario
was conducted  twice  for  each array  of the vertical strings-once
during morning hours  (10:00  to  11:15 a.m.) and once  during evening
hours  (4:00  to  5:15  p.m.).   The  afternoon experiment for type  2
was  repeated,  and the  data  from  this seventh experiment  were held
 in  reserve.   To  restore indoor CO concentrations to background
 levels  between morning and  evening experiments, windows  were
 opened  at 2:00 p.m.  and closed at 3:00 p.m., allowing an hour  for
 cessation of anj air movement patterns due to the window openings.

     A possible source  of interference for these experiments was
 local lutomobile traffic.  However, historical data from the
 research site indicated that brief outdoor transients coinciding
            e
witeary morning and late afternoon traffic peaks would be
relatively rare and on the order of 5 ppm or lower should they
occSr.  The extent of interference from any outdoor CO spikes was
quantified with the stationary monitoring network, through which
outdoor CO concentrations were measured every 15 minutes.
                                11

-------
                                 ODD
                                 D
                                       \
                                              DOWNSTAIRS


Figure 4.  General locations of the release point  (denoted by an
asterisk), the vertical sampling planes  (denoted by crosses joined
by broken lines), and the stationary monitoring network  (denoted
by filled circles).
                                12

-------
      Table 2.  Sequence of Contaminant-Release Experiments
  Date
     Release
     period
 Experiment type
(array of vertical
 sampling strings)
December l

December 2


December 3


December 4
 4:00 to  5:15 p.m.

10:00 to 11:15 a.m.
 4:00 to  5:15 p.m.

10:00 to 11:15 a.m.
 4:00 to  5:15 p.m.

10:00 to 11:15 a.m.
 4:00 to  5:15 p.m.
     Type 1

     Type 2
     Type 2

     Type
     Type

     Type
     Type 2,
                                13

-------
2.3.     AQCillary_Measurements

    The GEOMET research house facility contains a complete indoor
and outdoor network of continuous sensors for indoor environmental
conditions, operating status of major appliances, and
meteorological conditions outdoors.  The parameters selected to
assist in data interpretation included indoor air temperature and
the meteorological parameters of windspeed and direction,  solar
radiation, outdoor air temperature, and precipitation.   Siting
criteria for these parameters are listed in Table 3.

    In addition to environmental parameters, air infiltration
rates were measured using the tracer-dilution method (ASTM 1981),
with sulfur hexafluoride (SFg) as the tracer.  Before each
experiment, SPg was injected into the research house with an
automated system that provided rapid mixing through the central
forced-air heating and cooling system.  Following the period of
SFg injection, the air circulation system was kept off throughout
each experiment.  Indoor sampling locations for SPg were identical
to those used for the stationary CO monitoring network; a single
analyzer was used to sequentially sample each location, enabling
calculation of air infiltration rates over periods as short as
.15 minutes.  The instruments used to monitor SFg concentrations
and environmental parameters are listed in Table 4.


2.4.     Quality_Assurance and Control Procedures

    Data quality objectives were stipulated in the protocol
document (GEOMET 1987b) in terms of accuracy, precision, and
completeness of data collected during the experiments.  For most
parameters, the targeted accuracy and precision levels were
±10 percent; the targeted completeness of the data across all
measurement parameters and experiments was 95 percent.  Specific
procedures used to ensure the"collection of high-quality data
included external audits, multipoint calibrations, zero and span
checks, and additional routine activities performed by technicians
responsible for conducting the experiments and maintaining the
research houses.

    The most recent external audit at the research houses,
conducted during September 1987, involved (l) challenging GEOMET's
gas analyzers with known concentrations of National Bureau of
Standards (NBS)-traceable standard gases and (2) colocating
NBS-traceable instruments with GEOMET's meteorological and
indoor-environment sensors for parallel monitoring and comparison
of instrument responses.  Audit parameters of relevance to this
investigation Included the Beckman analyzer used for the
stationary CO network, four of the nine portable GE CO detectors,
the SFg analyzer, all outdoor meteorological sensors, and a subset
of the indoor-temperature sensors.


                                14

-------
  Table 3.  Specifications for Ancillary Measurement Parameters



Parameter
Windspeed
Wind direction
Solar radiation
Number
of
indoor
sites
0
0
0
Number
of
outdoor
sites
1
1
1



Location
10 m above ground3
10 m above ground3
Roof of house3
(total)

Air temperature
Precipitation
Air exchange rate
Indoor:  Centroid
of each major room";
Outdoor:  1.5m
above ground3

Gauge opening 0.3m
above ground3

Colocated with
stationary probes
for CO
a In general accordance with Section 3.0 of the EPA Quality
  Assurance Handbook f or Air Pollution Measurement Systems :
  Volume IV.  Meteorologi^4-...,JJ^MM.f5lf-S-^,  (EPA-6QQ/4-82-Q6Q) .
  In accordance with criteria established through previous
  experimentation reported in "Energy Use, Infiltration, and
  Indoor Air Quality in Tight, Well-Insulated Residences"  (EPRI
  Report No. EA/EM-4117), prepared by GEQMET for the Electric
  Power Research Institute,
                                15

-------
 Table 4.  instrumentation for Ancillary Measurement Parameters

  parameter      instrumentation   Manufacturer       Model

SF               Gas chromato-     S-Cubed           215BGC
                  graph/electron
                  capture detector
Temperature      Thermistor        Omega             OL-700
Windspeed        Anemometer        climatronics      WM-III
Wind direction   Vane              climatronics      WM-III
Solar radiation  Pyranometer       Matrix            MKI-G
Precipitation   ' Tipping bucket    Quallmetrics      6021A
                                 16

-------
    Multipoint calibrations were conducted for all gas  analyzers
at the beginning and end of the 4-day period during which
contaminant-release experiments were conducted,   in addition,  zero
anl spaSChecks were performed before each morning £P«^  fs
conducted, the results of which were recorded on control charts.
Based on historical performance data, control limits of ±0.75  ppm
at zero and ±1.5 ppm at span (9 ppm) were established for all  CO
monitors.  Any CO monitor exhibiting a response outside the
control limits was recalibrated before a new experiment was
initiated.

    On arrival at the research houses each day, the technician
first performed routine operational checks of various types of
eouipment.  Findings and observations were recorded on a daily
checklist form (Figure 5) to establish the degree of general
readiness for planSd activities.  The experiments were designed
to operate without intervention by the technician.  Principal
areas of attention included physically rearranging the detailed
network of CO analyzers, verifying the readiness of the analyzers
(based on zero and span checks) and data  acquisition systems  and
initiating/terminating the source release.  Measurements of air
infiltration and other auxiliary parameters proceeded through a
computer-controlled system with specified intervals for sampling
and recording instrument signals.

    Critical daily actions were recorded  on an operational
checklist  (Figure 6).  A typical dally schedule  involved the
following sequence of events:


          (1)  After making the types  of  operational checks
              described  above,  the  data  file  containing results
              from the previous day's experiments was  closed  and a
              new file was opened for recording  results from  the
              current day's  experiments.

          (2)  After verifying that  concentrations from the previous
               afternoon's  experiment had receded to acceptable
               levels,  the  technician reconfigured the  detailed
               network of CO analyzers as necessary and performed
               zero  and  span checks.

          (3)   source-feed  connections to the release room were
               checked and  the morning release was Initiated.   The
               technician reviewed the progress of the experiment
               by monitoring selected parameters on the CRT screen
               for the central data acquisition system.
                                 17

-------
  H'/AC  on-line:   14
15
      Gas       [  ]  C   ]
      Electric  (  J  [   ]
      Other     [  ]  [   ]
                                        Date;
                                        J.O.;  ~^~
                                        Time:  	
                                        Technician:
                           Daily Checklist - Instrumentation

     Performance Checks  (Verify reasonableness of data and check)

                                        parameters

                                   Room temperature
                                   Energy
                                   HVAC and duct temperatures
                                   RH and room velocity
                                   Pressure and velocity
                                   Furnace fan status change
                                   Furnace gas counter (verify operation)
                                   Pollutants  & meteorology

     Record screen display values  for  the following (during HVAC operation):

      14     IS                         14     15
14 IS Channels
[ ] [ ] AQ - C07
f
f
[
[ :
£ ]
[
f
*

c :

MM
CIO - CIS
DO — £03
KH — £09
£10 — PCS
£04 or 601
H03
10 — 115
                    AO ^,:SR  Temp)
                    84 (IR Tanp}
                           Oil (Supply Temp)
                           014 (Return Temp)
    Status/Logical  Checks  (Verify  status or parameter change after system begins operation)

     14     IS

                    Furnace fan
                    Furnace gas valve
                    Supply temperatures Increase
                    Furnace thermocouple
                    HVAC system initiated at appropriate room  temperature
    Instrumentation checks (verify operation, flows,  reasonableness of data-
    insert "1"  if operational or "0" if off-line)
                                    1
    [  ]   SFe  zone sequence        [  ]
           Pollutant zone sequence
           DAS  clocks
           SFg  Analyzer
           Pollutant Analyzers
           N0X/NQ A                    ]
           NOX/NO 8
           C02-A
           C02-8
                   C  ]   [  1   C   1    C   ]  RADON
                   [  ]   [  }   [   ]    [   ]  RADON PROGENY
                   Pressure sensors  [specify any off-line)
                   Halocarbons

                   CO-A
                   CQ-B
                   02
                   Other:
0.  Comments:
                      Figure  5,    Daily  checklist  form.
                                             18

-------
c.
           CHECKLIST—CONTAMINANT MIGRATION EXPERIMENTS

    Experimental  Configuration:

    Instrumentation

    GS CO Detectors

    Channel  Serial No.  Status

       1
Source Release Data

      Daytime
    Start
    Time:
                Stop
                Time:
        Plow:
    Tank Concentration

    	 % CO in Air

    Data Files

    Campbell Datalogger

    File Name; 	
    Start: 	
    End:  	

    Clock Status  [  ]

    Comments:
                                       Placement in
                                      Sampling Plane
7
8
9
4
5
6
I
2
3
Start
Time:
 Overnight

      Stop
	  Time:
                                     Flow:
                                 Tank Concentration

                                 	 % CO  in Air



                                 Site DAS

                                 File Name:	
                                 Start:  	
                                 End:                ~~
                                 Clock status   [  ]
Figure 6.   Operational checklist  for experiments.
                               19

-------
         (4)  After completion of the morning experiment,  the
              technician verified that concentrations  had  declined
              to acceptable levels and initiated the afternoon
              release of CO.  The. afternoon experiment then pro-
              ceeded to automatic shutdown of the CO release and
              continuing measurements under program control.


2.5.     Data Processingand Analysis Procedures

    All instrument signals were scanned, averaged, and recorded at
prescribed intervals by computer-controlled data acquisition
systems.  Instruments measuring meteorological parameters  were
scanned at l-minute intervals and recorded as hourly averages.
Measurements with the stationary sampling network for CO and SFg
were recorded at 3-mlnute intervals corresponding to the times at
which various locations were sequentially sampled.  A separate
data acquisition system was devoted to the network of nine CO
detectors to enable the recording of 1-minute averages at each
sampling site.  All instrument outputs were recorded as voltages
on iBM-PC-compatible diskettes.

    Calibration factors (slope and intercept) derived from
multipoint calibrations were applied to the raw data at GEOMET's
data center through programs implemented on IBM personal computers
and compatibles.  The calibrated data were reviewed for
unreasonable values such as negative concentrations and sharp
excursions from smooth trends  ^e.g,, a temporary decline of  one
data point to near-zero values during a period of otherwise  steady
growth).  Questionable values were flagged to alert analysts to
sections of valid and Invalid data.

    Data analysis efforts were keyed to the basic objective  of
these experiments—exploring the implications of contaminant
migration patterns for current methods of estimating human
exposures.  Fundamental avenues of analysis included  (1)
comparisons across experiments through graphical and statistical
methods, (2) comparisons of measured values with those predicted
by currently used models, and  (3) evaluations of model
refinements.

    Initial stages of data analysis focused on basic concentration
profiles as well as similarities and differences across different
experiments.  This comparative analysis was applied to CO
concentrations as well as to air infiltration rates, indoor
temperatures, and meteorological conditions.  In particular, the
spatial and temporal profiles of CO concentrations near the
doorway to the master bedroom were assessed for consistency  across
experiments; the vertical string of CO analyzers at this
                                 20

-------
location was the anchor point for linking results across the three
types of experiments.

    The functional relationships among the vertical strings
constituting the detailed CO sampling network and stationary CO
monitoring sites are illustrated in Figure 7,  To summarize and
integrate results across experiments, statistics such as peak and
average CO concentrations were used in additiqa to graphical
summaries.

    The general mass balance model currently used for indoor air
exposure assessments was applied to estimate Indoor concentrations
in the region of the source and at remote locations, based on
measured values of air infiltration rates, source release rates,
and outdoor CO concentrations.  The general framework for mass
balance models is summarized in Appendix A.

    Model residuals  (algebraic differences between calculated and
measured concentrations) were analyzed to identify model
assumptions or outdoor conditions that led to significant
differences as well as good correspondence.  Particular attention
was given to identifying conditions where the single-chamber model
provided poor estimates of passive or active exposures.

    The final stages of data analysis explored various avenues of
model refinement including multichamber models for estimating
passive exposures.  Results from previous PFT measurements at the
research house under similar outdoor conditions were included as
inputs to the multichamber model.
                                21

-------
(Corner Bedroom]
\   probe     /
             \
                                  * Release
                                    Point
              \
                \
                            S2
                   ANCHOR
                  (8|, Di. Hi)
              D
   {Front Bedroom!
       Prob©
H,
        ,   \    .
        Basement
         Prob»
H3
 i
 I
	  Mixing in Room of Release

.—	  Transport to Adjacent Rooms

rrr •—"•  Transport to Remainder of House

-„--  Associations Between Vertical
       Strings (indicated by S\, Of, H|)
       and Stationary Sites (indicated
       in parentheses)
                 ,         \
                 /Livina Area)
                 \  Prob«  /
  Figure 7.   Functional relationships among vertical  sampling
  strings of the detailed monitoring network and stationary
  monitoring sites.
                                  22

-------
3.       ANALYSIS OF EXPERIMENTAL RESULTS

    Outdoor conditions prevailing during each of the experiments
are summarized in Section 3.1, and data quality levels associated
with critical measurement parameters are summarized in Section
3.2.  Concentration profiles for each experiment are presented in
Section 3.3, and the results across all experiments are integrated
in Section 3,4.


3.1.     Sxpe r imental Conditions

    Sampling was conducted during the first 4 day.' of December
1987.  Three different configurations of portable CO detectors
were used In the master bedroom, hallway, and other bedrooms.  Two
experiments were performed for each configuration, one starting at
10 a.m. and another starting at 4 p.m.

    Both ambient and indoor conditions can influence the
concentrations and rates of migration of contaminants in indoor
environments.  Indoor temperatures, ambient carbon monoxide
concentration, air infiltration rate, and meteorological
conditions (winds, temperature, and solar radiation) were used as
a basis for summarizing the prevailing conditions during each
experiment.  Average values, standard deviations, and ranges for
these parameters during the 4-hour period of CO release and decay
for each experiment are given in Table 5.

    During each of the six experiments, both indoor and outdoor
temperatures remained within 10 percent of the mean value.
Although temperature differences between the morning and afternoon
experiments were observed, the overall temperature difference
between indoors and outdoors was quite similar across the pair of
experiments for each configuration of CO detectors.  Air
infiltration rates usually were slightly higher during the
afternoon than morning experiments.

    The average windspeed was typically as high or higher during
morning than afternoon experiments.  Solar radiation levels were
also higher during morning than afternoon experiments, mainly
because the afternoon decay period included hours after sunset.
The highest ambient CO concentration during any experiment was
1.4 ppm.  This relatively low level, coupled with upstairs air
infiltration rates averaging about 0.3 air changes per hour,
implies that outdoor CO concentrations had negligible impacts on
indoor concentration levels.
                                23

-------
Table 5,   Summary of  Prevailing Conditions During Each Experiment
Date/period
of experiment

Parameter
Part A: Experiment Type I—Vertical Sampling
Horning
(12/04/87)











Afternoon
(12/01/87)











Air temperature (*F)
Ambient outdoor
Master bedroom
Corner bedroom
Front bedroom
Hallway
Living room
Downstairs
Upstairs infiltration rate (h"
Ansblont outdoor CO (ppm)
Windspetd (tn1/h)
Wind direction (degrees)
Solar radiation (Btu/ft2)
A1r temperature (*F)
Ambient outdoor
Master bedroom
Corner bedroom
Front bedroom
Hallway
Living room
Downstairs
Upstairs Infiltration rate (h
Ambient outdoor CO (ppn)
Wlndspeed (nl/h)
Wind direction (degrees)
Solar radiation (Btu/ft2)

Average
Strings In

38.2
63.3
63.0
63.5
64.5
64.2
63.0
•1) 0.30
0.8
5,0
291
21

44.3
64.6
64.4
64.7
65.6
65.3
64.1
-1) 0.29
0,7
2.8
231
1
Standard
deviation

Range
Master Bedroom

0.7
1.3
1.2
1.3
1.3
1.3
0.9
0.02
0.1
0.6
11.5
5.4

0.6
1.3
1.3
1.3
1,1
1.2
0.7
0.07
0.5
1.1
22.1
2

37.6 to 39.1
61.9 to 64.9
61.7 to 64.6
62.1 to 65.1
63.2 to 66.2
62.8 to 65.8
62.0 to 64.1
0.28 to 0.33
0.6 to 0.9
4.4 to 5.8
278 to 303
14 to 16

43.9 to 45.2
63.3 to 66.2
63.0 to 66.1
63.3 to 66.3
64.4 to 67.0
64.0 to 66.8
63.3 to 65.0
0.23 to 0.36
0.3 to 1,4
1.8 to 4.3
211 to 258
0 to 4
                                                               (Continued)
                                24

-------
Table 5.   Summary  of  Prevailing Conditions During Each Experiment  (Continued)
Date/period
of experiment Parameter Average
Part B: Experiment Type 2— Vertical Sampling Strings
Morning
(12/02/87)











Afternoon
(12/02/87)











A1r temperature (*F)
Ambient outdoor
Master bedroom
Corner bedroom
Front bedroom
Hallway
Living room
Downstairs
Upstairs Infiltration rate (h'1)
Ambient outdoor CO (ppm)
Wlndspeed (ml/h)
Wind direction (degrees)
Solar radiation (Btu/ft*)
A1r temperature (*F)
Ambient outdoor
Master bedroom
Corner bedroom
Front bedroom
Hallway
Living room
Downstairs
Upstairs Infiltration rate (h"1)
Ambient outdoor CO (ppm)
Wlndspeed (ml/h)
Wind direction (degrees)
Solar radiation (Btu/ft2)

41.
67.
66.
66.
68.
68.

0.
0.
10.
310
95

36.
63.
62.
63.
64.
63.
62.
0.
0.
6.
333
2.
Standard
deviation
Range
In Hallway

7
9
3
6
3
0

30
3
2



6
6
6
0
g
8
3
36
7
6

5

1
1
0
0
0
0

0
0
2
52
35

2
2
1
1
1
1
0
0
0
1
3

.2
.0
,4
.4
.4
.6

.02
.2
.1
.2
.8

.0
.2
.8
.8
.7
.8
.9
.05
.2
.4
.5
5

40.
66.
66.
66.
67.
67,

0.
0.
7.
263
42

34.
61.
60.
61.
63.
61.
61.
0.
0.
5.
329
0

0
9
1
4
8
2

26
0
7



6
2
6
1
0
8
3
29
5
1



to
to
to
to
to
to

to
to
to
to
to

to
to
to
to
to
to
to
to
to
to
to
to

42.
69.
67.
67,
68.
68.

0,
0.
12.
357
121

39.
66.
64.
65.
66.
66.
63.
0.

7
2
0
2
7
7

32
4
2



2
3
7
2
9
0
4
42
1.0
8.2
337
10


                                                                    (Continued)
                                     25

-------
Table 5.   Summary of  Prevailing Conditions During Each Experiment (Concluded)
Date/period
of experiment
Paraneter
Average
Part C: Experiment Type 3— Vertical Sampling Strings
Morning
(12/03/87)






A1r temperature
Ambient outdoor
Master bedroom
Corner bedroom
Trent badrooB
Hallway
Living room
Downstairs
OF)







Upstairs Infiltration rate (h"1}
Ambient outdoor CO (ppm)

Wlndspeed («1/h)

Wind direction (degrees)

Afternoon
(12/03/87)






Solar radiation
A1r temperature
Ambient outdoor
Master bedroom
Corner bedroom
Front bedroom
Hallway
Living room
Downstairs
(BtU/ft2)
OF)







Upstairs Infiltration rate (h"1)


Ambient outdoor
Wlndspeed (m1/h)
CO (ppm)

Wind direction (degrees)

Solar radiation
(BtU/ftz)

41.
66.
65.
65.
66.
66.

0.
0.
6.
194
95

40.
61.
62.
62.
63.
63.
62.
0,
0.
6
175
1.
Standard
deviation
Range
1n Doorways of Adjacent Bedrooms

0
2
1
5
6
9

26
2
4



3
9
0
2

0.
0.
0.
o;
0.
0.

0.
0.
0.
7,
31.


8
8
5
6
6
6

04
3
4
1
3

1.8
1.4
1.5
1.3
,4 1.2
,0
,5
,31
.3
.6

.5
1
1
0
0
1
54
1
.3
.0
.03
.2
.2
.1
.7

40.
65.

0
1
64.5
64.8
65.
66.

0,
0,
5,
189
51

37
60
60
60
62
61
61
0
0
5
97
0
,8
,0

.22
,0
.9



.9
.4
.4
.8
.0
.6
.5
.28
.1
.2



to
to
to
to
to
to

to
to
to
to
to

to
to
to
to
to
to
to
to
to
to
to
to

41.
66.
65.
66.
67.
67.


7
9
6
1
1
4

0.32
0.
,5
6.8
205
123

42
63
63
63
64
64
63
0
0
8
214
4



.0
,6
.8
,9
.9
.7
.7
.35
.5
.2


                                      26

-------
3.2.      Data Quality

    Data quality levels associated with this investigation are
summarized in this subsection in terms of measurement accuracy  and
precision.  The focus is on the primary measurement parameter for
the investigation--concentrations of the surrogate contaminant
(CO)    Results from a recent external performance audit,  as well
as multipoint calibrations and zero and span checks co- -ring the
specific period of investigation, were used as inputs ^o  the
assessment of data quality for this parameter.  The performance of
measurement systems for air infiltration rates and meteorological
parameters is also characterized.

    During September 1987, an external performance audit was
conducted at GEOMET's research house facility by the Center for
Environmental Quality Assurance of Research Triangle ^titute
(RTI 1987).  Four of the CO detectors and the Beckman CO analyzer
used in this project were included in the audit.  Each analyzer
was challenged with zero air and with CO concentrations of 5, 10,
20  30  and 45 ppm.  The audit results are summarized in Table 6
in'terms of a regression equation expressing the relationship
between analyzer response and audit concentration.  Perfect
agreement would be indicated by a slope of one, an intercept of
zero, and a correlation coefficient of one.  As shown in the
table, slopes for all analyzers were within ±0.01 of unity, and
intercepts were relatively small in magnitude for all analyzers
except GE detector #108, which exhibited a negative bias of  1 ppm.
All correlation coefficients were either 1 or 0,9999.

    The accuracy and precision of the  analyzers at an audit
concentration of 10.34 ppm are summarized  in  Table 7.  The
accuracy  objective of 10 percent was met or exceeded  by  all
detectors.  Aside from GE detector  #108,, which had an accuracy
level of  -10 percent due to  the  1-ppm  negative bias,  the accuracy
of the detectors was within  3 percent.  The precision across
detectors was ±5 percent, well within  the  objective  of ±10
percent.  Thus, these results  indicate that the performance  of  the
CO analyzers was quite  satisfactory at the time of the audit.

    Accuracy and precision of  all nine CO  detectors  and  the
Beckman  analyzer used during the Investigation  are  shown in
Table  8,  based  on  responses  to  a final calibration concentration
of 9.06  ppm.  The  accuracy of the analyzers ranged  from  -2.4 to
+2.5  percent,  and  the  precision across all analyzers was ±1.4
percent.   Thus,  accuracy  and precision levels at  the final
calibration were  well  within data quality objectives of  ±10
percent.
                                 27

-------
Table 6.  Summary of External Audit* Results for Five CO Analyzers


                                   Regression of analyzer
                             response on audit concentration

                                                       Correlation
Analyzer                  Slope         Intercept      coefficient
GE detector #142
GE detector #119
GE detector 4108
GE detector #104
Beckman model 866
0.99
0.99
1.01
0.99
1.00
-0.12
-0.07
-1.01
0.16
0,26
1.0000
0.9999
0.9999
1.0000
1.0000
  Audit conducted on September  21,  1987.
      Table 7.  Accuracy  and  Precision of Five CO Analyzers
                at Audit Concentration of 10.34 ppm
Analyzer Response, ppm
GE detector #142
GE detector #119
GE detector #108
GE detector #104
Beckman model 866
Average, all analyzers
Standard deviation
precision, percent
10.10
10,24
9.31
10.37
10.79
10.16
0.54
±5.3
Accuracy, percent
-2.3
-1.0
-10.0
+0.3
+ 3.1



                                28

-------
    Table 8,   Accuracy and Precision of all CO Analyzers Used  in
     the Investigation at Final Calibration3 Input of 9.06  ppra
Analyzer
GE detector #037
GE detector #104
GE detector #108
GE detector #119
GE detector #123
GE detector #130
GE detector #142
GE detector #147
GE detector #153
Beckman model 866
Response, pprn
9,12
9.07
8,84
9.11
8.87
9.00
9,07
9.29
9.11
9.11
Accuracy, percent
+ 0.7
+0.1
-2.4
+0.6
-2.1
-0.7
+0.1
+2.5
+0.6
+0.6
Average, all analyzers       9.06

Standard deviation           0.13

Precision, percent          ±1.4
a Final calibration conducted on December 7, 1987.
                               29

-------
    Changes in slopes and intercepts for the CO detectors between
beginning and ending calibrations are summarized in Table 9,
Changes in slopes were minimal--? percent at most and 2 percent or
less for six of the nine detectors.  Drift in the intercepts  was
somewhat more pronounced; eight of the nine detectors exhibited a
downward drift in the intercept, averaging 0.5 ppm.  This downward
drift was most likely due to the sensitivity of the CO detector to
temperature; as indicated in Section 3.1, indoor temperatures
during the experiments were between 60 and 70 "F because the
central heating system was turned off during the conduct of each
experiment.

    Intermediate zero and span checks were performed on 36
occasions (nine detectors on 4 days each); in seven of these
cases, detectors drifted to an out-of-control state.  However,
four of these cases were associated with a single detector (#108).
In contrast to the beginning and ending calibrations that were
performed in the laboratory adjacent to the house, zero and span
checks were performed Inside the research house so that the
detectors would not be removed from the testing environment.
Bags filled with zero and span gases were used for the zero and
span checks, as opposed to multipoint calibrations that fed the
gases directly to the detectors from cylinders.  This practice
resulted in a slight negative bias in detector response to the
zero and span checks, as illustrated for one detector in Figure 8.
Thus, because virtually all detectors exhibited a downward drift
between the beginning and ending calibration and because the use
of bags for zero and span checks resulted in a negative bias in
detector response, the control limits used as criteria for reca-
libration were stricter than intended.  Thus, detectors may have
been recalibrated in selected cases when calibration was not
necessary.  However, the only Impact of extra calibrations, if any,
would be a minor improvement in accuracy.

    The analyzer used to guantitate SFg concentrations for
calculation of air infiltration rates was also Included in the
external performance audit.  Based on audit concentrations ranging
from zero to 800 parts per billion (ppb), the regression of analyzer
response on audit concentration resulted in a slope of 0.94,  an
intercept of 7;2, and a correlation coefficient of 0.999.  The
average difference between audit concentrations and analyzer
responses was 5.5 percent.  Based on initial and final calibrations
surrounding the period of investigation, instrument drift was
negligible; Initial and final slopes were 1.02 and 0.99, and
initial and final intercepts were 4.1 and 6.5.  At a calibration
concentration of 400 ppb, the midpoint of the analytical range of
the instrument, accuracy was +1 percent for the initial calibration
and +2 percent for the final calibration; precision at the 4QO~ppb
concentration was ±1 percent.  Thus, the performance of the SFg
analyzer also was well within data quality objectives.


                                30

-------
Table 9.  Changes in Slopes and Intercepts Between Beginning and Ending
            Calibrations for CO Detectors Used in the Investigation
Beginning calibration
(11-30-87)
Detector
t037
1104
#108
1119
1123
*130
1142
#147
1153
Slope
0.99
0.97
0.98
0.99
.1.01
0.98
1.01
0.99
0.99
Intercept
0.26
0.48
0.13
0.11
0.10
0.45
-0.11
0.41
0.04
Ending
(12
Slope
0.99
0.97
1.05
0.92
0.99
0.99
1.05
1.01
0.99
calibration
-7-87)
Intercept
0.13
-0.23
-1.21
0.36
-0.59
-0.21
-0.41
0.26
-0.14
Change in
slope, percent
0.0
0.0
+7.1
- 7.1
-2.0
+1.0
+4.0
+2.0
0.0
Change in
intercept, ppm
-0.13
-0.71
-1.34
+0.25
-0.69
-0.66
-0.30
-0.15
-0,18

-------
   E
   a.
   a
   o
   o
   c
   o
   Q

   o
   o
I U
9_
8 —
7 -
5 ™"
4 -
3 -
2 -
1 -
0 -
-1
	 	 ^ . ...
In!
Caltbi
tlal
•atlon


». 	 —
"--— ^
Interrr
Che<


^_ — f
ilttent
;ks




^-"
Fin
Calibr
al
ation
[ _ 	 -a


'it ii 	 i 	 	 1
123456
                          Sequential Check Number

                     Zero Check          o   span Check
Figure 8.  Response of one CO Detector to 0  and  9.06 ppm during
multipoint calibrations and daily zero and span  checks.
                                 32

-------
    A representative subset of thermistors used to measure room
air temperatures was also included in the performance audit.   Of
the 12 room thermistors audited, 8 differed from the audit reading
by less than 0.5'F, two differed by less than 2°F, and 2 failed to
meet data quality objectives.  The two thermistors that failed
were immediately replaced, prior to this investigation.  Subse-
quent checks have verified proper performance of all thermistors
in use at the research houses.

    All meteorological instrumentation exhibited satisfactory
performance during the audit except the windspeed/direction
sensor.  Windspeeds measured by the onsite sensor were low due to
bearing wear.  The unit was returned to the manufacturer for
maintenance and calibration prior to this investigation, and
additional quality control procedures were implemented to verify
proper performance.


3,3.     ConcentrationProfiles

    This section summarizes basic patterns from the detailed
monitoring network that reflect mixing and transport of the CO
tracer.  For each experiment, the monitoring points in the
sampling plane were first treated as three vertical strips to
examine trends in the vertical, and were then recompared by
viewing the data as three layers to examine horizontal trends.


3.3.1.   Room of Release

    In the room of release, tracer concentrations at all six
points began to rise rapidly very soon after tracer release was
initiated and continued to climb as the tracer release was
terminated (the end of the release period, 1.25 hours after the
start, is indicated by a vertical line in Figure  9 and in
subsequent figures).  Peak concentrations were reached shortly
after the source was turned off and then declined smoothly.

    As shown in Figure 9, transport through the open doorway
to the anchor string in the hallway was fairly rapid; concen-
trations at the anchor string rose and fell on essentially the
same timing as in the room.  Concentration gradients persisted
during the release.  At vertical string 82 (between the release
point and the doorway), concentrations were highest near the
floor.  At vertical string 83, concentrations were highest near
                                 33

-------
o
      Morning  Experiment
          Anchor String





E
a
a





14-
12 -

10-

8-

6-
•
4-
»
2
n -







.-***
^**
* * "
,* .**I«**"
. :.•

m Top

+ Mid

o Low

*
*
• .***..
.:. :?...**»».
»iSS|js±ft**.
•iiitSiiii;

                                       Evening  Experiment
                                            Anchor String
  1       2       3
 Elapsed Time, Hours

S2 String of MBR Plane
14
-
12-
-
10-
-
S 8
a
a 6.
,
4-
2-
n -








»*
****
* *
* «"
.:,»*:***

• Top

* Mid

o Low

* *
*

»%
SiiSSiSSSt! i f

                                                123
                                                Elapsed Time, Hours

                                              S2 String of MBR Plane
       1
 Figure 9.   Vertical concentration gradients with CO detectors
 arrayed  in  the master bedroom.
14
12

10
8-
6-

4-
2 -
0-



*
* *
_ *
**
* fl**
* *
* |B
m9
m*
* « *"
*
p *



t
^
,t.
a *
«*
1 *


' I ™««— ,*«•— — — — — — — ^-y™™™™™,,^ 	 j, 	
14-
la-

la-
s'
6-

4-
•
o-



*
•***
*
« j
* »••"
»*
% *'
%»*
*





•
***$
** j
"i||!Sl!Ms...,,



01234 01234
S3 String of MBR Plane S3 String of MBR Plane
14-1
12-
10-
8-
6-
4-
2-
o-


."*
,*"""S»$:J

* *
«*




*
**'•,
l||:s:-.
"""'*::»M::;:;J»»..
•»..
i 	 i 	 1 	 1 ' i
14-
12-
10
8-
6-
4-
2-
0-
r




.."!
, ,f!^*'
."I*11"
.**




*r?
%:';:,.ii
"::***:::I;;-*».,M,,
*•••»«..
I 1 O *? A
                                34

-------
the ceiling.  At the anchor string, the highest concentrations
occurred at midlevel.  This pattern may be due to convective
transport in the master bedroom.  The temperature network,
however, is insufficiently detailed to fully quantify such
effects on thermodynamic grounds.

    When the data were recomposed to form horizontal layers
(Figure 10), the situation became clearer.  Concentrations in the
middle layer (1.2 meter height) at the two strings in the room and
at the anchor string in the hallway were remarkably similar, rising
and falling 'in unison and showing essentially the same pattern for
the morning and evening experiments.

    Convective circulation in the room of release could lead to
uneven transport in the room of release, creating and sustaining
concentrated packets that are slowly mixed into the general
volume.  In the top and bottom layers, where highest concentrations
occurred in the room during tracer release, concentrations were as
much as 40 percent higher than in the middle layer.  Once the
source was turned off, concentrations were rapidly equalized in
the room.  Vertical gradients, however, were sustained at the anchor
string in the hallway for at least an hour after the source was
turned off, and substantial differences persisted in the top and
bottom layers between the room of release and the anchor point.

    Although concentration profiles in the room of release appear
to be driven by convection, the pathways of transport within the
room cannot be fully appreciated because the sampling pl»ne only
provides a two-dimensional section through a three-dimensional
transport field.  Many different plume configurations can be
envisioned that would lead to the same concentration profiles.
Nonetheless, the following insights can be drawn from the
perspective of mass balance modeling;   (1) mixing proceeded rapidly
in the horizontal and slowly in the vertical under conditions of
natural air motions; and (2) in the room of release, convective
motions created and sustained concentration gradients that led to
poorly mixed conditions while the source was active.  Concentration
gradients in the room of release dissipated rapidly once  the  source
was turned off.


3.3.2.   Transport to Adjacent Rooms

    Concentration profiles for the vertical sampling strings
erected  inside the doorways of rooms  that  adjoin the room of
release  are shown in Figure 11.  Vertical  gradients were  largely
dissipated by the time contaminants reached the D£ string in  the
corner bedroom and the 03 string in the front bedroom.  Peak
concentrations were substantially lower than in the room  of
release, and concentration profiles were delayed by approximately


                                35

-------
      Morning  Experiment
          Top  L©vel
Evening  Experiment
    Top L®v@t
E 8-
CL
a 6.

4-
2
ft
.•
» 4
•«*,«^»
» .
**
* «.
* I*1
«
,:il"

.*
*
*
* *'»*
"""""I:U"«..(M|
*****»»

E 8
Q.
o- 6-
•
4-
2-
n ,


*»*
.t*»
**
ft m
**« B
*vl'""""
«*"


1*.
« .
9
»
'-""•.*-Js»*M
*^8i5||||B|««illl

01234 "6 i 2 3 4
Elapsed TIm», Hours Elapsed Tim®, Hours
Mid L®v«l Mid LOVQI
14-
12-
10-
8
-
6
4-
2
0 -




»
,:** *
- I2*'''
.»'



+ *
****
"•*,



14-
12-
10
8

6
4-
2-
0





.•*;**
»a"j***
•*»
i*




*+
"!|ii»
*|:!I!:»:«-MMM.I


0123401234
Low Level Low Laval
14-
12-
10-
8-

8-

4-
2

o


*»*
**
* .«*

•f 41
/ ."•"""

	 , 	 ,.,,,, —


*
».*
*»

**.




14-
12
10-
8-

6-

4-
2-

« -


^
^^*
+**
*
# ^
« *»* »"

«*••*""


*
*
*
*
** *
*• *.
• **
,«ii»»»"*»i»jjjjj
***•••»»,
***«•»»
Figure 10.  Horizontal concentration gradients with CO detectors
arrayed in the master bedroom.
                               36

-------
     Morning  Experiment

        Anchor String
Evening  Experiment
   Anchor String
8

6
4
2
0
f- j, *
« *
"

.
-I ai


*
*
*•.
t::::il';;i«t...4Mttt


E 8
Q.
Q. g
4
2
n


.-"*'*
** t .r
.** .*/ .
\* .:/"'
i m*

*
*
•"I***.
M!l|||«««««a,MHl

V •:..,,, U -, 	 , 	 , 	 , 	 , 	 1 	 , 	 , 	 1
012340123*:
Elapsed Time, Hours Elapsed Time, Hours
D2 String In CBR D2 String In CBR
14-
12-
10-
8-
6-
4-
2 •
0 -





.•'
«*** « J
* * *l
...•«*




..
""'"'''''"•••'••••••I,,,

14-
12-
10-
8-
6-
4-
2.
f\






,tr,(»i«|t|tf





„«•'..,„- . '
••.....
U T • ...... y T 	 , 	 1 	 • 	 , 	 , 	 ,_ 	 , 	 , 	 1
0123401234
D3 String In FBR D3 String In FBR
14i
12-
10-
8
6-
4-
-
2-
0 •






Ma«ii*!'S:
*.** — . . .






.*»**'•""'**, *«»
"""•••*».l;;;j;j;;
..
14-
12
ID-
S'
6-
4-
B
2-
— . i







,tns:i
*#«ii6*






»********-,»#
********«**»»»

0 1 2 3 4~6'i"2'3'4
Figure 11.  vertical concentration  gradients with CO detectors
arrayed in adj acent bedrooms,
                              37

-------
30 minutes and flattened.  For the evening experiment,  concentra-
tions in the bottom level at the anchor string were higher than at
the mid and top levels; this was the only experiment that deviated
from the basic pattern of higher concentrations at the  midlevel
anchor.

    Significant horizontal gradients persisted in the middle layer
between these two rooms and the anchor point.   As shown in
Figure 12, midlevel concentrations were higher at the anchor for
as long as an hour after the source was turned off.  During the
morning experiment, concentrations in the lower layers  of the two
bedrooms rose and fell in unison, suggesting strong coupling in
the lower layer and rapid mixing within the two rooms.

    During the evening experiment, horizontal gradients between
the bedrooms and the anchor point were sustained at all levels for
nearly an hour after the source was turned off.  Nonetheless, the
concentration profiles in both bedrooms were very similar to each
other and to profiles from the morning experiment.

    The following insights can be drawn from these two experiments:
(1) contaminant transport into the adjacent bedrooms was primarily
through the lower layers; and {2) mixing within these rooms was
fairly complete and rapid.


3.3.3.   Transport Through the Hallway

    Concentration profiles for the vertical sampling strings
erected along the hallway are shown in Figure 13.  Vertical
concentration gradients were evident for essentially the entire
4-hour period at the % string  (midway down the hall) and at the H3
string (just past the foyer).  By the time the contaminant
transport reached the H2 string, the highest concentrations were
found near the floor.  Concentrations at the H2 string were
consistently lower than at the anchor string.  At the H3 string,
concentrations were even lower and more nearly homogeneous in the
vertical.

    Regular pulsations were strongly evident in these concentration
profiles that did not prevail in the two bedrooms, supporting
convective coupling of the hallway and living area, possibly
including the basement zone through the stairwell.  Recomposing
the data to layer form (Figure  14), it is .apparent that  for the
evening experiment the oscillations in the middle layer  after the
                                 38

-------
  Morning  Experiment

        Top Level
Evening  Experiment

     Top Level
14-
12^
10"
E 8-
o.
6 •
4-
2
o-
C



y\
..,ii#':!''
• Anchor
+ D2
o D3
**•*,.
"»»'»,„„„,
14-
12
10-
E 8-
Q.
a 6.
4-
2-
n .



•
„•""".
••• ***!*.*
„;.**»***
• Anchor
+ D2
o D3
'"ill"
•::-:!I!I-««.M,,WHIMMJ
' 	 1 	 1 	 ' 	 1 	 1 	 	 1 	 1 	 1 	 1 W • ~ "• 	 T- 	 - T ••- I 	 - I 1
3123401234
Elapsed Time, Hours Elapsed Time, Hours
14-
12
10-
8 -
6-
,
4-
2-
o-
C
14-
12-
10-
8-
6-
4-
2
0-
(
Mid Level



. >.*"
«
s*
•' ..*••'
• ^*
••****



a
*
m
*
* *+ 1 1 J y 9 • « ^ _
) 1 2 3 *
Low Level





.«:«'
.;;!J.-:!t





|llfi"" 	 M|||||M
11111111
) 1 2 3 <
14
12-
10-
8-
.6-

4-
2-
o-
1- (
14
12
10
8
6
4-
2-
0-
1 (
Mid Level





*
••••'!!!*««i:i




•
;
• n
31234
Low Level




.'"""
.»
.*:»*
" »..*s:*
*»«,»*'



m
m
•
•
•
"'"""*""*«»•
D 1 2 3 4
Figure 12.  Horizontal concentration gradients with CO
detectors arrayed in adjacent bedrooms.
                           39

-------
          Morning  Experiment


              Anchor Siring
Evening Exporlmont

    Anchor String
E
a.
a.
14-
12 -
1 *»
1O •

8

*
6 -

4-
2-







^ *
+ +
*
*
•**

,iB-«
" Top

* Mid

* Low
«•»*


*
*
.»„***»„
8*
-------
      Morning  Experiment
           Top L@v@l
         1
         Elapsed Time, Hours

             Mid Lovol
            Low L@val
14
12
10

 6
 4
 2
 0
                                            Evening  Experiment
                                                  Top L®V®I
14-
12
10-
£ 8-
a.
a 6.
4.
2-
0 •
	 —






......I'1"'''
a Anchor
+ H2
o H3



.Ji»
:..:!ii'"«"«w«MMMI)lmil
^ <% 4 j*
14
12-
10
I •;
«J
4-
•
n



*
m
.V*
jS."*'**1""
a Anchor
+ H2
o H3

' r?ll-.-»

14-
12-
ID-
S'
6-
4

2

0
(



§.
.•*
*
m
»*"
«
***

»*• ...
ti*.j:* *



« Ba
a
a
a
.**>»+** *'"B«»a

„***** t"|flf if 9


3 ' 1 2 3 4
             §.*«.,
            *.«         '•*
                         *»«»»•.«*»,
                                             1       2
                                            E1ap««d Time, Hours

                                                 Mid Lovol
14
12
10
8
6-

4-
2
0
(



0

tt»
a "•»
• •
B
* ******
liiltttt 	 t




••..
BB
*""•••
r*****i***-*.ii!;t 	 ••«.....
•«••• * 	 ••». *tt:*-' *****
* •••»•«•...
31234
                                                 Low Laval
14-
12
10
8
0-
4

2
0
<





* BJ**
•»**
• * **
••• *




..
cn»_
"•••.....
t **•*«***•-*.»*».,»,."***'***;•

3123^
Figure  14.   Horizontal  concentration gradients with CO detectors
arrayed  in  the hallway.
                                41

-------
source had been turned off were parallel  and  extended  from the
living area (H3^ to at least midway down  the  hall  (H2),  but were
not reflected in the middle layer at the  anchor string.   |hese
oscillations were faintly reflected in tne top layer at  the anchor
and at Ho, suggesting general downturn.  Strong convectiye action
reflected in the middli and top layers at the anchor string was not
coupled to profiles at the H2 or H3 sites.

    During the morning experiment, concentration profiles were
much smoother.  Although some oscillations were evident, vertical
gradients were dissipated fairly slowly but smoothly at allthree
lites, while horizontal gradients in the bottom and top layers
dissioated within 30 to 45 minutes.  In the middle layer,
horizontal gradients between the hallway sites largely disappeared
5?tiiS an hour, but did not equilibrate with the anchor position
until nearly 2 hours after the source had been turned off.

    The following general insights can be drawn from these
experiments:


              Concentrations in  the hallway were lower  than at
              the anchor  site;

              Although vertical  gradients were weaker in the hall,
              some  stratification existed with highest
              concentrations prevailing  near  the floor;  and

              Convective  motions In the  hallway provided fine
               structure  to  the concentration profiles.


 3.4.      integration Across Experiments

     To view experimental results in terms of active and passive
 exposures,  unification of the  three basic experimental  types  is
 needed for synthesis of a general case linking the room of
 release,  adjacent rooms,  and the remainder of the  indoor volume.
 The main elements that enable  this integration of results are the
 repetitive features of the experimental design.

     in addition to strict repetition of time-related elements such
 as release rate and duration,  the experimental design featured a
 stationary monitoring network to measure CO ^entrations at
 three heights at the anchor site, at midlevel heights at ^ur
 other indoor locations, and outdoors.  As previously illustrated
                                 42

-------
a
in Section 2.5 (Figure 7), each vertical sampling plane provided
a series of intermediate sampling sites connecting the anchor
point to each of the fixed monitoring points on the main floor ot
the research house and to the room of release.

    The research house itself represents perhaps the most
powerful point of integration across the experiments in that it
is a realistic full-scale model responding to changing
environmental conditions; thus, the primary differences among
validated data from experiment to experiment are traceable to
naturally occurring changes in transport and mixing patterns.
Although this features does not necessarily lead to a simple
equivalence for uniting all experiments, it nonetheless presents
additional information on the range of variability that prevails
under real-world conditions.

    Data integration across experiments was carried out from
three perspectives:


              For each height level at the  anchor location;

              At the midlevel height for the anchor location  plus
              the stationary network;  and

              At the midlevel height for the anchor,  stationary
              network, and mobile  sampling  locations.


    Concentration profiles at the  anchor point *are  compared  across
all experiments in  Figure 15.  At  the  midlevel  of the anchor,
which was strongly  associated with events  in the room of  release
as well  as in the adjacent rooms,  peak concentrations varied  by
nearly a factor of  2  across  all  experiments but  rose  and
fell in  very  similar  fashion.  This pattern indicates that
similar  forces wers at work, but at different  intensities.   The
highest  peak  concentrations  in the middle  layer were  associated
with morning  experiments  when  convective  coupling would be
assisted by solar gains.  During the concentration decay  period
following the end of  CO  tracer release, concentrations at all
levels converged  to a fairly narrow interval.

    When attention  is shifted  to the mid-height of the anchor
string  in relation  to the rest of the  stationary monitoring
network  (Figure  16),  it  can  be seen that  additional factors came
 into play.  Well-mixed conditions were usually approached within
 an hour  after the release was  turned off,  but  concentration
                                 43

-------
                         Top L@v@l Anchor String©
             E
             Q.
             Q.
14
12
10
8
6
•
4-
2-
n J




X
**
** * IV
.**•«!$•"
«ii**°"
" AM1
+ AM2
o AM3
& PM1
x PM2
It f. V PM3
»l'l*lIl*St It *****«*»«-.


                            1       2       3


                           Elapsed Time, Hours
                         Mid Level Anchor Strings
14
12
to
a



*.x;.
» B«~

4-
•
2-
n -
a"**!1*
•s?Vr*
***»"**

»**



«
i* *«
T**« "+
*T*'****th»».«
*"»*»* Mf 35 ****»»»*
*ff*'M||$|«jjjj*jjj

                         Low Leva! Anchor String*
                  14

                  12^

                  10

                   8

                   6

                   4

                   2
                   0
                                    2 •
Figure  15.   Concentration profiles across  all experiments at the

anchor  point.
                                  44

-------
      Morning Experiment
         Master Bedroom
Evening Experiment
    Master Bedroom
         1       2      3

        E!apsod Tlmo, Hours

             Hallway
                                                   Out
                                                   Down
                   CBR
                   FBR
                   Anchor
   123
   Elapsed Tim®, Hours

       Hallway
12
           Other Bedroom
    Other Bedroom
Figure  16,   Concentration  profiles for each experiment at the
midlevel anchor point in the hallway and stationary monitoring
sites.
                                 45

-------
profiles were sometimes affected by transport to the basement and,
during some of the evening experiments, by infiltration of out-
door CO concentrations from mild outdoor peaks due to local traf-
fic.  There generally was greater separation between upstairs and
downstairs concentrations during morning than evening experi-
ments.  Concentrations during the release period generally rose
much more quickly at the anchor site than at any other location.
The single exception to this trend was an evening experiment
during which concentrations in the corner and front bedrooms rose
near"!*  in unison with those at the anchor.

    To develop characteristic concentration profiles connecting
all measurements taken in the middle ."  -er of the main floor, the
data sets were combined in three stage=».  Firs^, site-specific
averages across all experiments were constructed for each location
that was common to all experiments (i.e., rnidlevel anchor and
the stationary network).  Second, data from the midlevel of the
vertical sampling strings that were relocated from experiment to
experiment were normalized as a fraction of the corresponding
midlevel anchor concentration at the end of the release period
for each experiment.  Third, the normalized values were
multiplied by the grand-average midlevel anchor value at the end
of the release period to rescale the normalized data in terms of
average conditions.

    Concentration profiles obtained through this exercise were
examined as iS-minute snapshots.  Throughout the release period
(Figure n), concentrations in the room of release and at the
anchor point in the nearby section of the hallway were virtually
identical.  Concentrations tapered off substantially beyond the
anchor point, such that noticeable increases in the other
bedrooms and living room were not evidenced until 45 to 60
minutes after the release was started.  Mixing within the
adjoining bedrooms appeared to be rapidly achieved, as indicated
by very similar concentrations near the doorway and center of
each room.

    Following the release period (Figure 18), spatial uniformity
throughout the house was nearly achieved as concentrations in the
release area receded and concentrations in other areas continued
to rise.  In particular, within 60 minutes after the release
period ended, concentrations began to recede in all upstairs
locations that were monitored, and concentrations in likely
receptor locations  (each bedroom and the living room) were within
0.6 ppm of one another.  Further unification of experimental
results and contaminant migration patterns can be obtained
through modeling efforts presented in the next section.
                               46

-------
15 ninetis »ft«r
                                                   30 mlnutai ifter reli4s« surted
                                     UPSTAIRS
                                                        .-cm
                                                      Mtitir EtdrooB

                                                        3.5
    u
                                                                              Kttchffi
                                                               3,8   , 0,5      0.1
                                                                         /\
                                                              v
                                                              \a.s
                                                      Corntr
                                                      Bodrooa
                                                       1.2
 Frant
lldrgoa
                                                                   1.2
                                                                          Entry
                    Olntr.j a oar.
                                                                                       1.2
                                                                                        ROM
                                                                                         U?STAIBS
            r*liiM ttirttd
                                                     Blnutti iftir must lUrtttt
                                      UPSTAIRS


                             75 minutiJ »ftir rtltist ittrttd
                             («nd of rtlillt p«rlod)
	 [,.,. 1,1, —
Hiitir Btdroc
6,9
sx \/ y
1.4
Corntr
Bidrooa


_ ^<4
'•M«^J 	 '

x
\k
U
bx
' V
X
M
/\

Bsdrooa
2.7






5"
6
1
i
-I" 	 ' 	 ' 	 ' 	
y KUcnin
I dining Utan
Z.3 Z.I

Entry




Living ROM
i — t — i — i 	

UPSTAIRS
 Figure 17,   Spatial  profile  of CO  concentrations  (in  ppm)
 upstairs  during  the  release  period.
                                                                                         UPSTAIRS
                                               47

-------
15 minutes after relent ended
                                          30 nlrvytej after release ended
i ' — *" — ' ^4r-*~~^" 'S
7.0 H U
/\ A ^^v
J 	 J 6.5 ,. 3.1
v \s y y^
lA £
x \/ v
\3.2
Corner Front
Sedroom Bedroom
3.7 2,9
t lnl 	 t 	 J — [ — ! 	


s/ \J[;S Kitchen
njfjj
_0
1.9
Entry



_JLi
Dining Room
2,9
Living Room
[_"f 	 | 	 ; 	



UPSTAIRS


3.8
' 1 , , ,, 1 3-7
v v' / ,
3-\ 1
V
Corner
Sedroon 8
3.5

[=3^ /
"In
tsJ^
/ 2-S
~V NX '
3
Front
edroon
3,3


\ f? Kitchen
DfM
Jj
1.8
s\
Entry


1 Dining Roon
2.9
Living Root*
UPSTAIRS
   Figure  18.   Spatial profile of  CO concentrations (in ppm)
   upstairs  following the  release  period.
                                      48

-------
4.       ANALYTICAL MODELING

    The practical objectives of the work reported here involved
(1) examining advantages and limitations associated with
simplified mass balance modeling of complex scenarios^ and
(2) identifying avenues of useful improvements througn
multichamber modeling.  Modeling activities began with the
generalized single-chamber mass balance model, using inputs of air
exchange, source rates, and indoor volume.   Multichamber modeling,
using interzonal airflows from previous PFT measurements analyzed
by Brookhaven National Laboratory (BNL), was then carried out to
identify improvements resulting from this next level of model
complexity.
4.1.
Single-Chamber Mass Balance Model
    The single-chamber mass balance model (see Appendix A) was run
on a 15-minute time step to calculate indoor concentrations at
time intervals consistent with averaging periods for the
stationary monitoring network.  Separate calculations were
performed at air exchange rates of 0.2, 0.3, and 0.4 air changes
per hour  (ACH) to cover the range of air infiltration rates that
occurred during the experimental period.  All model calculations
assumed perfect mixing and negligible outdoor levels.

    A3 shown in Figure 19, peak concentrations under nominal
conditions coincide with the end of CO tracer release and occupy a
fairly narrow concentration interval, ranging from 3.2 ppm for the
0 4-ACH case to 3.6 ppm for the 0.2-ACH case.  By the end of the
experimental period, the dilution effects of the different levels
of air exchange decrease concentrations by half while broadening
the differences between cases.

    The single-chamber model, when applied to the general air
volume of the main floor of the research house, significantly
underpredicts peak concentrations observed in the room of release
and at the anchor point.  Figure 20  illustrates this lack of
correspondence for a morning experiment.  In this figure,
concentration profiles from the middle levels at the anchor and
S->/S3  (master bedroom) strings are plotted together with model
estimates calculated at 0.3 ACH.  The  calculated peak
concentration of  3.4 ppm falls short of measured values  by  a
factor of 2,  and  the model does not  approach measured values
until near the end of the decay period.
                                 49

-------
£
Q,
a

c
o
c
0)
o
c
o
o

o
o
        0
1            2

    Elapsed Time, Hours
 Figure 19.   Single-chamber mass balance  model calculations for

 generalized experimental conditions.
                                 50

-------
£
Q.
Q.

tf
O
c
CD
u
c
O
O

O
O
   Anchor

+  S2

O  S3

   Predicted
                          Elapsed Time, Hours
 Figure 20.  Comparison  between single-zone model  predictions and
 measurements near the release area (experiment  type  l,  morning
 run).
                                 51

-------
    Comparisons at monitoring points that are distant  from the
release area, however,  begin to show remarkable agreement.   The
concentration profiles  for distant strings of the hallway plane
(Figure 21, upper part) and the strings inside the doorways of
rooms adjacent to the room of release (Figure 21, lower part) are
largely reproduced by this simple model.

    The primary shortfall of the model for areas distant from the
release point lies in the relative timing of peak concentrations.
Modeled peak concentrations occur at the end of the tracer-release
period.  At the midlevel of the hallway,  the measured  peak at the
H2 string, midway down the hall, occurred 5 minutes after the
release ended.  At the far end of the hall (the H3 string), the
peak occurred 30 minutes after the release had ended.   At the D2
and D3 strings, located Just Inside the doorways of the second  and
third bedrooms, measured peaks occurred 10 minutes after the
release ended.  On the day of this experiment, ambient levels  of
CO were at approximately 1 ppm as the experiment began; although
the outdoor concentration declined as the experiment progressed,
the outdoor influence resulted in a slight offset that was not
reflected in tne model.

    Figure 22 illustrates the correspondence between the nominal
model case and 15-minute measurements from the stationary network.
Although the model estimates at these points indicate an earlier
occurrence of the peak concentration than do the measurements,  the
general correspondence is excellent.

    The single-chamber mass balance model does not account for
spatial gradients; it tracks the overall retention of the
contaminant, providing estimates that correspond to volume-
weighted averages.  To examine this concept, volume-weighted
average concentrations were calculated on a  15-minute basis using
data from the stationary network and the midlevel probe  of the
anchor string.  As shown in Figure  23, these volume-weighted
averages are in good agreement with the nominal  model case.  Even
though concentrations near the release are up  to 4 times higher,
they occupy only 20 percent of the total volume.  The single-
chamber mass balance, then, provides a fairly  close approximation
of general concentration profiles that relate  to passive exposure,
but tends to underestimate concentration profiles that relate to
active exposure.
                                 52

-------
£
CL
Q.
C
O
Q
C
O
O
O
O
15
14
13
12
1 1
10
 9
 8
 7
 6
 S
 4
 3
 2
 1
 0
     15
     14
     13
     12
     11
     10
      9
      8
      7
      6
      5
      4
            Hallway
                                         Anchor
                                         H2
                                         H3
                                         Predicted
                          Elapsed Time, Hours
Other Bedrooms
                                              •  Anchor
                                              +  D2
                                              O  D3
                                             —  Predicted
Figure 21.  Comparison  between single-zone model predictions and
measurements in the hallway and front bedrooms  (experiment types
2  and  3,  morning runs).
                                  53

-------
  E
  a
  a.
  c"
  o
  c
  3
  o
  c
  o
  o
  o
  o
15
14
13 H
12
11 -
10 -
 9
 8
 7 _
 6 -
 5 -
 4
 3 ~
 2 -
 1 II
 0
 «  LR
 +  CBR
 O  FBR
—  Predicted
                           Elapsed Time, Hours
Figure 22.  Comparison between single-char.iber model predictions
and measurements  at  stationary monitoring locations that  represent
likely sites of passive exposures (experiment type 2, morning run).
                                 54

-------
E
a
a.

c
o
c
o
o
c
o
o

o
o
     14 -
     12 -
10 -
      0
                                 Average Indoor Concentration


                                 Predicted Concentration
                      1             2

                          Elapsed Time, Hours
Figure 23.  Comparison between, single-chamber model predictions

and  volume-weighted  average indoor  concentrations for  three

morning experiments.
                                   5S

-------
    The single-chamber model cannot simultaneously  estimate  active
and passive exposures with equal accuracy because the emissions
-re dispersed (in the model) to the general indoor  air volume and
-re therefore equal throughout the house.  Prom these experiments,
>L is obvious that there was both confinement in the room of
release ^lus time-consuming transport to other rooms, resulting  in
the strong spatial and temporal differences that were observea.

    in concept, the single-chamber model could be exercised  twice,
first using the volume of the release room to estimate active
exposure and then using the general volume to estimate passive
exposure.  However, shrinking the reference volume to that of the
release area (master bedroom) would result in overestimates  of
active exposure.  For example, for these experiments  substituting
the 40-m3 volume of the master bedroom lor the 215-mJ volume of
the main floor would increase peak modeled concentrations by the
ratio of the volumes (5.4), producing peak concentrations near
20 ppm where approximately  10-pprn levels were observed; this dis-
crepancy is due to the dilution effects  of transport to other
rooms.

    There are two alternatives for increasing prediction
accuracy—(1) developing empirical values to increase the
effective removal   ite of contaminants when using  a  single-zone.
model to estimate active exposure and  (2) using  a  multichamber
model to simultaneously estimate active  and passive  exposures.
The second  alternative—using  a mulcichamber model—would seem  to
be a  more natural and straightforward  approach.


4.2.     Multiple-Chamber Modeling

    A two-chamber mass balance model was formulated  using
interzonal  flows derived  from  previous  PFT measurements  at  the
research house.  For this model  analysis,  the  master bedroom was
defined  as  zone  1  (volume of 40  m3), and the  living/dining  area,
hallway, and the two smaller bedrooms  were treated as the second
zone  (volume of  175 n»3).  Specific  equations  used  in tMs analysis
are presented  in Appendix A.   For  this synthetic case,  flow
coupling between the upstairs  and  the  basement was ignored.  .As
shown previously  {Section 3),  CO tracer would be occasionally
transported to the  basement zone,  but  this  occurrence did not
significantly  alter  the  general form of the  concentration profiles
upstairs.
                                 56

-------
    The system of interzonal and infiltration/exfiltration flows
used in the two-chamber model is illustrated in Figure 24.  These
airflows were directly adapted from PFT measurements that were
taken over an 18-hour period in January 1985 at the research house.
This particular set of PFT measurements was acquired to assess
baseline conditions for experiments being conducted at that time.
As with the experiments reported here, operation of the central
circulation fan was suppressed.  Interzonal airflows derived from
the PPT data, although not necessarily equal to those that
occurred during the experiments reported here, nonetheless provide
characteristic values that are appropriate for model applications.

    The two-chamber mass balance model provides substantial
improvement in estimating active exposure, while retaining the good
correspondence with passive exposures that was obtained with the
single-chamber model.  As shown in Figure 25, calculated peak
concentrations reached 9.2 ppm in the master bedroom (zone 1) and
3.0 ppm in zone 2.  Model calculations reproduce measured
concentration profiles in the midlevel at both the anchor site and
the far end of the hallway for the morning experiment when the
vertical sampling plane was placed along the hallway.

    Even though environmental conditions varied across
experiments, the two-chamber model calculations, which are
predicated on nominal conditions, provide good estimates of peak
and average concentrations for all of the experiments.  Table 10
summarizes the range of measured peak and average concentrations at
key indoor locations, along with estimates from the single- and
two-chamber models.  The single-chamber model is in best agreement
with measurements in the living area, and increasingly
underestimates peak and average concentrations at locations that
are closer to the release area.

    For the two-chamber model, calculated peak and average
concentrations for zone 1 are in the center of the observed range
for the six analyzers situated in the master bedroom for one of the
morning experiments.  At the anchor site and other locations where
measurements are available for all experiments, the two-chamber
model passes through a transition phase where zone 1 estimates are
overtaken by zone 2 estimates in terms of agreement with measured
values.  The anchor site represents perhaps the most important
transition between the two defined chambers.
                                57

-------
             5.7 mi / h
          38,5 m3 / h
     3.1 m3/h
               f
              74.5 nf / h
                                    77.1 nf /h
                                             41.1 nf /h
       Zone  1
       Master Bedroom
       Volume = 40 m3
Zone 2
Remainder of House
Volume - 175m3
Figure  24.   Airflows used as inputs  to a two-chamber model,
                                   58

-------
£
Q.
a
E
a.
a.
     15
14
13 H
12
11 -
10 -
 9 -
 8 -
 7 -
 6 -
 5 -
 4 -
 3 ~
 2 -
 1 -
        0
                    Zone  One  Concentrations
                                Midlevel at anchor
                     1             2
                         Elapsed  Time, Hours
                    Zone  Two  Concentrations
                                MldlQvel at H2
                         Elapsed Tims, Hours
   Figxire 25.  Comparison  between two-chamber model predictions
   (indicated by a line) and  measured concentrations in each  zone
   (indicated by squares)  for morning experiment type 2,
                                    59

-------
  Table 10,  Comparisons o" .iodel Estimates and Measured Values
         for Peak and Time-weighted Average Concentration
              During and After Contaminant Release


                            Peak                 4-h Average
                        concentration           concentration
                            (ppm)
MODEL ESTIMATES

Single chamber               3.4                    2.2

Two chamber—zone 1          9.2                    4.5

Two chamber—zone 2          3.0                    2.1




RANGE OF MEASURED VALUES

Master bedroom            (7.1-11.0}a           (3.7-5.0)a

Anchor site—midlevel      6.3-8.2               3.8-4.7

Corner bedroom             3.9-4.5               2,3-2.9

Front bedroom              3.7-3.8               2.3-2.7

Living room                2.9-3.6               2.1-2,7
a Range of values from six sensors in one experiment; all others
  are range of values across all experiments.
                                 60

-------
    In the two-chamber model, the anchor site would be assigned to
the general air volume (zone 2) because it receives material from
the release zone and is physically outside the master bedroom.
However, as shown in Section 3, concentration profiles at the
anchor site differed from the convective patterns in the master
bedroom and from the well-mixed conditions that were approached in
other rooms.  Developing a separate volume centered on the anchor,
supported by correcting flows to direct the transport to other
rooms, would be an interesting avenue of possible refinement to
the two-chamber model.  Although such an approach would offer
refinements in reproducing time-related events such as the delay
to reach peak concentrations away from the release point, the
physical justification for defining additional volumes requires
additional information.
                                 61

-------
5.        DISCUSSION

    An overall perspective relating to goals and needs for
assessment of consumer exposures is provided in Section 5.1.
Major insights gained from this investigation and future actions
that can be taken to fill specific types of information gaps  are
discussed in Section 5.2.


5.1.     General Perspective and Needs

    An important objective pertaining to OTS exposure assessments
is to provide accurate estimates of active and passive exposures
resulting from different patterns of consumer product use in
residential environments.  One means of meeting this objective is
by developing a computer model that can provide accurate
predictions of exposures for a wide variety of scenarios.

    As indicated in Figure 26, a number of user inputs, table
lookups, and computer calculations are needed to support a
fully-specified computer model for active and passive exposures to
emissions from consumer products used in residential environments.
Many of the lookups and calculations go beyond the capabilities of
the Computerized Consumer Exposure Models (CCEM) in current use
but could be obtained through future data acquisition efforts.

    in Table 11, vital inputs to model lookups and calculations
are shown in relation to various types of data sources that
currently exist or that could exist as a result of future data
acquisition efforts.  Four possible sources of data—results from
chamber studies, consumer surveys, measurements with
perfluorocarbon tracers  (PFTs), and experiments in research
houses--are Indicated in the table; in addition, surveys in which
contaminants released from consumer products are monitored in
residential settings will also be needed to. validate consumer
exposure models.  Alternative approaches for filling information
gaps through chamber studies, consumer surveys, and monitoring
surveys were discussed in a recent report (GEOMET, 1987c) that was
prepared for EPA.

    One of the most significant voids in current exposure
assessments is the lack  of appropriate values  for air  infiltration
rates and interzonal airflows.  As shown in Table 12,  combining
data from past PFT measurements and future  research house
experiments can substantially  fill this gap.   These two  types of
data sources are very complementary;  PFT measurements  cover a
                                 63
     Preceding page blank

-------
                  Inputs from Model Users:

                Type of product and application
                Location and duration of use
                Type of structure
                Geographic area
                Time of year
                Type of heating/cooling system
                Status of windows/exhaust fans
                                          f
     Model Lookups;

  Structure/room volumes
  Time-varying emission
  and decay rates
  Outdoor meteorology
           Model Calculations:

         Air infiltration rates
         Extent of operation of
         heating/cooling system
         Interchamber airflows
         Time-varying
         concentrations of
         contaminant(s)
                                          f
                       Model Outputs:

                    Hourly and average
                    concentrations per
                    zone/room
                    Data files, tables, or
                    graphs
   Time-varying
Receptor Locations
Exposure
Profiles
      Figure 26.  Overview of consumer exposure model
                            64

-------
   Table 11.   Important Inputs to Model Lookups and Calculations
                    and Associated Data Sources
                                           Data sources
                                                            Research
                            Chamber  Consumer    PFT         house
      Type of input         studies   surveys  data base  experiments


 Emission rate

      Duration/rate of use              X
      for various  products

      Emission rate specific   X
      to  product and usage
      pattern

 Decay rate                    X

 Volume/mixing/transport

      Configuration and/or
      involvement  of
      floors/rooms

      Extent  of heating/
      cooling system
      operation

 Air  exchange

   -   Infiltration rate        X                 (X)           (x
                                                 \._x           ^^
   -   Ventilation  rate                  X                      fx

      Interchamber airflows    fxj                (xj


 X  =  Some data already exist.
'""x
 X)=  Data could be obtained  through future acquisition/assimilation
     efforts.
                                65

-------
Table 12.  Utility of PPT Data Base
    and Research House Experiments
  Characteristic
     PPT
  data base
                                     Research house
                                      experiments
Structure types and
geographic areas

Spatial coverage
per structure

Weather conditions
per structure

Internal conditions
per structure

Resultant airflows
Various
1 to 4 zones
Limited
Largely unknown
Average
                                  Limited
                                  Individual rooms
                                  Various
                                  Various (can be
                                  controlled)

                                  Time-varying
                                  or average
                 66

-------
 broad  array of geographic areas and structure types, but are
 limited  in spatial, temporal, and operational details for each
 structure; by comparison, research house experiments provide much
 greater  detail on temporal and spatial variations for various
 conditions of product use, but may have less generalizability.
 However, the transferability of results from research house
 experiments can be greatly aided by replicating selected
 experiments in other common types of occupied structures, such as
 apartments, townhouses, and selected configurations of
 single-family detached homes.

    As noted in a recent scoping report (GEOMET 1987a) on the topic
 of room-to-room contaminant migration, a relatively rich
 repository of data concerning time-averaged air infiltration rates
 and interzonal airflows for a variety of structures and geographic
 areas is currently maintained at Brookhaven National Laboratory,
 The analytical potential of these data cannot yet be exploited
 because the results are not fully unified in a computer-accessible
 format, but efforts toward this end have begun.  Such efforts,
 coupled with future research house experiments, will significantly
 reduce the current information void that must be filled to
 quantitate contaminant migration rates and model the exposure
 implications of such migration,


 5 • 2.     Insights from the Current Investigation

    The specific scenario used for this preliminary investigation
 involved a controlled point release of a surrogate contaminant
 over a period of 1,25 hours.   Although this scenario does not
necessarily represent any specific product or contaminant, the
 experimental results have a number of implications for passive
 exposures and modeling thereof.   Contaminant concentrations were
 generally 3 to 4 times higher in the room of release than in areas
where passive exposures could occur, but concentrations in the
 other areas were dlstinguishably above background levels within
 1 hour after the release was initiated.  Further, spatial
uniformity in concentrations throughout the main floor of the
 research house was approached within 45 to 60 minutes after the
 end of release period, during which time concentrations near the
 release area receded and concentrations in other areas continued
 to rise.  Interestingly, a single-chamber model—similar to that
 currently used In OTS assessments of active exposures In
 residential environments—more closely approximated passive than
 active exposures; use of a multichamber model resulted in better
 estimation of each type of exposure.
                                67

-------
    Through detailed spatial and temporal monitoring  the
 research house experiments described and analyzed in this report
 have provided substantial insights into short-term contaminant
 migration patterns, but were restricted to one release scenario,
 Further experiments are needed to address the following types and
 conditions of simulated product use:


         •    Type of release—point versus area {e.g., wall or
              floor);

         *    Location of release;

              Duration of release;

         *    Status of interior doors (open or closed);

         »    Status of windows and exhaust fans; and

         *    Season and operation of heating/cooling systems.


Covering all possible combinations of these conditions would
require a prohibitively large number of experiments; however, all
combinations are not needed to obtain vital insights.  The number
of experiments could be substantially reduced, for example, by
limiting the next round of investigation to (I) a standard type,
duration, and location of release, but under different conditions
relating to interior doors, windows, exhaust fans, and
heating/cooling system operation, and (2) varying types,
durations, or locations of release for a single set of conditions.
Results from a recent EPA-sponsored consumer survey (Westat 1987)
can be used to help determine the most common conditions
surrounding product use.

    To facilitate the conduct of experiments for a broader set of
simulated product usage scenarios, a detailed stationary
monitoring network needs to be established in the research houses;
this can be accomplished by expanding the number of monitoring
sites at the sacrifice of vertical detail at each site.  An
illustrative array of the CO detectors used In this investigation
for such a network is given in Figure 27; monitoring sites in the
bedrooms, living room,  and downstairs living area represent
probable receptor locations, whereas sites in the stairway and
hallway areas represent likely pathways of contaminant migration.
                                68

-------
                                                    UPSTAIRS
                 Garage
\
                                   Q
\
 Unfinished
Living Area
                                                   DOWNSTAIRS
PL.     27,   Illustrative monitoring array for  future contaminant
mir--    i  experiments  (filled  circles represent  probable receptor
lc-      »s;  empty circles represent likely migration routes).
                                  69

-------
    Providing a linkage between these experimental results and the
PFT data base is vital to the ultimate goal of improving the
accuracy and generalizability of consumer exposure models.
Consequently, future experiments should include PFT measurements
and real-time multiple tracer measurements as a standard
component.  An illustrative array of the four available types of
PFT sources is given in Figure 28; this particular array will  1)
enable quantitation of average airflows among the bedroom, living
>-oom  and downstairs areas and (2) provide a means of assessing
whether flows between bodroom arid living room areas can be
inferred from upstairs-downstairs flows (important because, in
many cases for multistory structures, PFT sources are configured
to provide estimates of flows between but not within stories).

    The transferability of results from research houses to a
variety of housing types can be significantly aided by replicating
selected research house experiments in the following types of
structures:

              Single-story, slab-on-grade structure;

              Single-story structure with basement;

              Multistory above-grade structure  (attached  and
              detached); and

          »    Apartment unit.


Structures inhabited by colleagues or  acquaintances could be  used
for pretests of this approach  in  a limited number  of settings.
^s little as one structure of  each type can  provide substantial
insights  regarding not only the transferability  of research  house
results but  also the need  for  broader  surveys  of this  type.
including a  detailed temporal  and spatial monitoring network  in
each house and making simultaneous PFT measurements will
strengthen the linkage between the detailed  research house results
and the time-averaged PFT  results that are  available for  a greater
variety of structure types.
                                 70

-------
                Bedroom Area

                 © ©
Living Room Area

   (2}  (T)
                   Garage
                                     Downstairs
                                     Living Area
Note:  Source types (3)and (4) represent a commonly used array of
       PPT sources for estimating flows between upstairs and
       downstairs areas; addition of source types  (T) and  (2}
       enables estimation of flows among three areaS--bedrooms,
       living room, and downstairs—and can be used to assess
       whether flows between bedrooms and living room could be
       properly inferred if only source types (T)  and (T)  had
       been used.                             ^^^
      Figure 28.  Proposed array of PFT sources for future
      contaminant migration experiments.

-------
6.        CONCLUSIONS AND RECOMMENDATIONS

    Major findings and conclusions stemming from the study
results and modeling efforts given in Sections 3 and 4 are
summarized in Section 6.1.  Recommendations drawn from the
discussion given in Section 5 are outlined in Section 6.2.


6.1.     conclusions

    The major findings and conclusions from this preliminary
investigation are as follows:


         *    During the 1.25-hour period when contaminant
              release from a consumer product in the master
              bedroom was simulated, resultant concentrations
              were 3 to 4 times higher in the area of the
              release than in other upstairs areas of the
              house.

              Within 45 to 60 minutes following the end of
              the contaminant-release period, concentrations
              throughout the upstairs of the house approached
              spatial uniformity, even though the central
              air circulation fan that would have promoted
              contaminant migration was kept off as part  of
              the experimental design.

         *    Vertical gradients in contaminant concentrations
              were most pronounced and variable in the release
              area and along the migration path in the hallway,
              suggesting that a fairly complex and somewhat
              variable system of forces is involved in the
              mixing and transport of contaminants.

         •    Concentrations at potential passive exposure sites
              such as the living room and bedrooms adjacent to
              the release area were generally similar in
              magnitude, even during the release period;  greater
              variations were observed during afternoon than
              morning experiments, possibly due to changing
              forces at play around sunset.

              Some evidence of contaminant migration downstairs
              was observed during selected experiments; however,
              even in these cases, downstairs concentrations were
              substantially lower than those upstairs.
       Preceding page blank        73

-------
             Application  of  a one-zone model similar to that
             used  for  current OTS exposure assessments resulted
             in  good estimates of passive exposure but
             substantial  underestimation of active exposure.

             Application  of  a two-zone model resulted in good
             estimates of both active and passive exposures
             upstairs; thus, even though a complex set of  forces
             may be involved in  contaminant mixing and
             transport, the  concept  of treating  general airflow
             patterns  as  a steady-state condition into which
             consumer-product emissions are  injected  and
             transported  appears valid and useful for improving
             exposure  estimates.

             The success  of  the  two-zone modeling effort  in
             estimating exposures  for the  release scenario
             examined  under  this investigation suggests that  PFT
             measurement  results coupled with  continuing
             experiments  in  research houses  will provide  a means
             of substantially  improving  exposure estimates for a
             variety of scenarios  relating to  use of consumer
             products  in  residential environments.


6.2.     Recommendations

    A broaderarray of experiments should  be conducted to
investigate contaminant migration fordifferent types of
releases and surrounding conditions^   Scenarios studied  at the
research house  should be expanded in terms  of (1) the type,
location, and duration of release and (2)  the status of interior
doors, windows, exhaust fans, and heating/cooling system,   A
detailed stationary monitoring network  should be established for
these experiments and measurements of time-.arying and
time-averaged interzonal airflows should be included as a routine
component of the monitoring design.  Such experiments will
improve our understanding of contaminant migration patterns and
exposure implications  for the greater variety of release types
and surrounding conditions that prevail in residential settings.

    The method of point release used for the current
investigation should be repeated with the new monitoring design.
A point-source release of 1.25-hour duration should be repeated
in  the master bedroom, first with the central air circulation  fan
off at all times and then with the fan on at all times.  These
two conditions should  then be repeated with the  release point
moved to the living room  and then downstairs.  Next, a different
                                74

-------
type of release (e.g.,  from an area source such as a wall or
floor) should be performed under similar conditions to evaluate
commonalities in migration patterns across different release
types.  Following this  sequence of experiments, the release type
and location should remain fixed but other conditions (e.g.,
interior doors, windows, and exhaust fans) should be varied,
first one at a time and then in selected combinations.  Finally,
selected types of experiments should be conducted with a consumer
product in use—one for which detailed monitoring of the
associated contaminant(s) can be performed at a reasonable cost.

    Selected types of research house experiments should be
replicated in a limited number of other structures to aid in
transferring the research results to variousresidential
settings^The number of scenarios should be restricted to four
at most, such as two release locations for each of two surrounding
conditions.  Other types of structures should include
single-story detached residences with and without a basement, a
multistory above-grade residence, and an apartment unit.  The
detailed monitoring network used at the research house should be
temporarily relocated to these other residences for this phase of
research.

    As soon the the BNL data base of PFT measurement  results has
been unified in a computer-accessible format,  analytical efforts
should be initiated.  The range and distribution of air
infiltration rates in different types of structures in different
geographic areas and at different times of the year should  be
determined from the data base.  Based on this  analysis,
characteristic airflows should be estimated for specific
structure-area-season combinations as inputs to future modeling
efforts.  The analysis should also assess whether  any systematic
relationship exists between the magnitudes of  air  infiltration
rates and internal airflows.

    In parallel with the research efforts described above,
activities to refine and improve currently used exposure
assessment models should be initiated.  This process  should  begin
with the development of a generalised multichamber model and
continue with refinements and expansions, as critical Inputs are
obtained from the research efforts recommended above.
                                75

-------
7.        REFERENCES


ASTM   1981.  Standard Practice for Measuring Air Leakage by the
T'-acer Dilution Method,   No. 6741-80.   Philadelphia,
Pennsylvania:  American Society for Testing and Materials.

GEOMET   1987a.  Scoping and Feasibility Study;  Room-to-Room
Contaminant Migration and OTS Indoor Air Exposure Assessments.
Report No.  IE-1820.  Germantown, Maryland:  GEOMET Technologies,
GEOMET   I987b.  Preliminary Experiments to investigate
Contaminant Migration:  Sampling and Analytical Protocol,  Report
No. IE-1807A.  Germantown, Maryland:  GEOMET Technologies, Inc.

GEOMFT   1987C.  Assessments of Human Exposure to Toxic
Substances in Residential Settings:  Alternatives for Data
Collection Designs.  Report No. IE-1826.  Germantown, Maryland:
GEOMET Technolgies, Inc.

RTI   1987   Performance Audit for GEOMET's Indoor Environmental
Program   Report No, RTI/3965/00-01F.  Research Triangle Park,
North Carolina:  Center for Environmental Quality Assurance,
Research Triangle Institute.

Westat  inc.   1987.  Household Solvent Products:  A  National
Usage Survey.  Report No. EPA-OTS  560/5-87-005.  Washington,
D.cT:  U.S.  Environmental Protection Agency, Office  of  Pesticides
and Toxic Substances.
        Preceding page blank       77

-------
                         Appendix A

            INDOOR AIR QUALITY MODELING CONCEPTS
                      AND FORMULATIONS*
The contents of this appendix have been excerpted from  "Scoping
and Feasibility Study:  Room-to-Room Contaminant Migration and
OTS Indoor Air Exposure Assessments," GEOMET Report No. IE-1820,
submitted to the Office of Toxic Substances, September  1987.
       Preceding page blank
79

-------
    The most widely used models for calculating contaminant
concentrations indoors represent the airspace of interest as  a
single well-stirred chamber or as a series of interconnected
chambers.  Rather than map the three-dimensional velocity and
dispersion field, these models track the amount of the contaminant
in the chamber(s) in terms of the mass balance defined by
generation; by inflow and outflow; and, for reactive contaminants,
by removal to sinks.

    Because the mass balance approach incorporates important
physical factors and processes directly, it has become the main
theoretical framework for indoor air quality simulations.  The
general mathematical expression of the mass balance is in the
form of a differential equation that, in solved form, constitutes
a user-implemented model.  Such implementations include
code-intensive computer programs featuring numerical techniques
{Nazzaroff and Cass 1986) as well as simplified approaches
involving analytical solutions applied through programs on desk-
top computers (Nagda et al. 1985).

    Single-chamber models define a given air space (e.g., a room,
a group of rooms or zone, or an entire building) as a single
well-mixed volume.  To extend the single-chamber approach to
multiple cnambers for quantifying zone-to-2one migration, the
indoor volume must be represented as a network of interconnected
chambers.  Contaminant mass balance is carried out for each of
these chambers; communicating flows with other chambers
(Figure A-l) are also considered.

    Because conditions in a given chamber are determined by
interactions with all other connecting chambers, the rnultlchamber
model is stated as a system of simultaneous equations.  The
mathematical framework for the multichamber description has  been
reviewed by Sinden  (1978) and by Sandberg  (1984).  General
equational forms for the single- and multichamber models are
presented  in Table A-l.

    From an operational standpoint, the most difficult decisions
concern appropriate model scenarios in terms of chambers and
airflows.  The general patterns  are of three basic types:  (1)  air
exchange between chambers and outdoors  (Qio, Qoi),  (2) chamber-
to-chamber airflows  (Qij), and  (3)  air circulation within  chambers,
These patterns are  Illustrated  in  Figure A-2.  The definition of a
chamber can entail  (1) the entire  building,  (2) a zone or  group of
rooms,  (3) a single  room, or  (4)  a  part of  a  room.
                                 81
       Preceding page blank

-------
TRANSPORTED
OUTPUT TO
OUTDOORS
                  Qlo
TRANSPOR
OUTPUT TO
OUTDOORS
         ED-*-
                 Qjo
                         INDOOR
                        SOURCES
                           SI
 INDOOR
SOURCES
   SJ
                                                            TRANSPOSED
                                                            INPUT FROM
                                                            OUTDOORS
TRANSPORTED
INPUT  FROM
OUTDOORS
          Figure A-l.  Basic mass balance  relationships
                    for multichamber approach.
                                82

-------
                         Table A-l.  Single-Chamber  and  Multichamber  Model  Summary

dCin
V 	
Single-chamber
r~* 4- n * r** 4* o * c* *
— o T vQi^out violin
Multichamber
dCj_ n n
V-i Gi + S 0-i i C4 S
a i_ „,—"»•« — t _* *

QijCi
           at
                                                    dt
00
CO
        INPUTS

           G = Source release rate  (g/h)
           V = Volume (m3)
         Qoi = Flow from outdoors  (m-Vh)
         Qj_0 = Flow to outdoors  (mVh)
             = Outdoor concentration  (g/
        AIR MASS BALANCE
         Qio = Qoi
INPUTS

 Gi = Source release rate in ith chamber (g/h;
 Vj_ = Volume of ith chamber (m3)
Q-H = Flow from jth to ith chamber (m3/h)
Q|^ = plow from ith to jth chamber (m3/h)
 Cj - Concentration in jth chamber (g/ni3)

AIR MASS BALANCE

 n         n
    Sr\ • •  —  p  n * j
    U^i -  *  Vij
        OUTPUT
        cin ~ Indoor concentration (g/ra3)
                                                OUTPUT
 i - Concentration in ith chamber  (g/m3)

-------
Sutdoors
Chamber i
Chamber j
1,   Air Exchange
    «  Natural Infiltration
    •  Natural ventilation
    •  Mechanical Ventilation
    •  Local Exhaust
                                             2.  Chamber-to-chamber Airflow
                                                 « Convective Circulation
                                                 • Advective Circulation
                                                 • Mechanical Circulation
                                              3.  Local Circulation
                                                 * Convective Mixing
                                                 * Mechanical Mixing
    Figure A-2.
        Basic patterns of air motion to be considered
                 in modeling.
                               84

-------
    Local circulation relates to the completeness of mixing.  In
the single-chamber description, effective volume and mixing factors
are sometimes employed to refine concentration estimates (Nagda
et al. 1987).  Within the multichamber description, incomplete
mixing signals a need for designation of additional chambers.  For
residential structures, this primarily involves defining zones
versus individual rooms.  Relatively little quantitative work has
been reported that would lead to general rules for subdividing
individual rooms to accommodate vertical stratification.

    Expanding the model perspective to the multichamber
description allows the exposure analyst to consider active
exposure and passive exposure simultaneously at the relatively
minor cost of additional complexity in calculations.  The primary
needs for implementation center on defining attributes of the
exposure scenarios that relate to volumes and flows.


REFERENCES

Nagda NL, Koontz MD, and Rector HE,  1985.  Energy Use,
Infiltration, and Indoor Air Quality in Tight, Well-Insulated
Residences,  EPRI Report No. EA/EM-4117.  Palo Alto, California:
Electric Power Research Institute.

Nagda NL, Rector HE, and Koontz MD.  1987.  Guidelinesfor
Monitoring Indoor Air Quality.  New York:  Hemisphere Publishing
Corporation.

Nazzaroff WW, and Cass GR.  1986.  Mathematical modeling of
chemically reactive pollutants in indoor air.  Environ. Science
Technol.  20:924-34.

Sandberg M.  1984,  The multi-chamber theory reconsidered from the
viewpoint of air quality studies.  Build.Environ. 19(4):221-233.

Sinden FW.  1978.  Multi-chamber theory of infiltration.  Build.
Environ.  (13):21-28.
                                85

-------