United States
               Environmental Protection
               Agency
               Research and Development
 Atmospheric Research and Exposure
 Assessment Laboratory
 Research Triangle Park NC 27711
 EPA/600/S3-91/013 Mar. 1991
EPA       Project  Summary
               Two  Indoor Air Exposure
               Modeling  Studies:  CONTAM
               Modeling  Results,  and Serial
               Correlation Effects
               Ronald E. Stogner, John S. Irwin, William B. Petersen, Mourad Aissa and
               Azzedine Lansari
                Indoor exposures have become an
               Important issue in recent years. It has
               been suggested that human exposures
               to pollutants indoors may significantly
               differ from outdoor pollutant exposures.
               There are, however, many factors that
               affect indoor exposures, such as local
               meteorology and building characteris-
               tics, and it is often difficult to isolate
               the individual factors affecting Indoor
               pollution using only measurement data.
               To  aid in quantifying  these  factors,
               therefore, numerical modeling studies
               are sometimes used. In this report we
               have undertaken such a study to better
               characterize some of the factors influ-
               encing indoor pollution and hence in-
               door exposures.
                Two mass balance computer models,
               CONTAM87 and CONTAM88, were used
               to estimate Indoor pollution levels for a
               hypothetical four-story building ex-
               posed to measured outdoor ozone,
               concentrations for a 24-hour period. The
               building contained 50 rooms;  exterior
               rooms contained windows,  Interior
               rooms did  not.  Modeled indoor con-
               centrations indicated that ozone levels
               generajly  decreased along a  partial
               cross section from peripheral rooms to
               interior rooms. CONTAM88 results in-
               dicated, however,  that  under certain
               conditions some peripheral rooms ac-
               tually  experienced lower ozone con-
               centrations than  interior  rooms.
               CONTAM88 analysis also revealed that
               simple actions such as opening interior
               office doors could significantly change
               indoor ozone distributions.
  Comparisons between CONTAM88
 and a simple one-cell box model indi-
 cated that the box model underesti-
 mated building-averaged ozone con-
 centrations  when compared  to
 CONTAM88 results using the same air
 exchange rates. Both models predicted
 changes in average ozone concentra-
 tions proportional to changes in  air
 exchange rates for the building.
 CONTAM88 maximum ozone concen-
 trations  were  less responsive  to
 changes in air exchange rates than
 corresponding CONTAM88 mean con-
 centrations.
  Statistical  methods of estimating
 concentration values for use In human
 exposure estimates have become  in-
 creasingly popular because of the
 complexities in correlating the temporal
 and spatial concentration variations
 within microenvlronments with the  lo-
 cations of people. The number of vari-
 ables and their associated uncertainty
 make deterministic models difficult to
 use. SHAPE and NEM, which include
 statistical methods for cha'racterizing
the concentration values for different
 microenvlronments, are  examples  of
 human exposure models that combine
deterministic  and statistical method-
ologies. In the simulations conducted '
thus far, these models  have ignored
serial correlation effects and therefore
 have tended to underestimate maximum
exposures. The purpose of the sensi-
tivity study conducted was to quantify
the factors affecting serial correlation
in indoor microenvironments. Further,
we Investigated  in a very preliminary
                                                            Printed on Recycled Paper

-------
way the  use  of personal exposure
monitoring data to Infer the values of
variables needed to estimate  Indoor
concentrations, such as the rates of air
exchange, pollutant  removal,  and pol-
lutant generation.
  This Project Summary was developed
by  EPA's Atmospheric Research and
Exposure Assessment Laboratory, Re-
search Triangle Park, NC, to announce
key findings of the research project
that Is  fully documented In a separate
report  of the  same title (See Project
Report ordering Information at back).

Introduction
  The  quality of air within buildings has
become an important issue in recent years.
In particular, it is recognized that outdoor
pollutant concentration levels may not ac-
curately reflect pollutant exposures expe-
rienced by persons indoors. Since  most
people spend a majority of their time in-
doors,  especially sensitive  groups  such
as  the  young  and the elderly, assess-
ments  of indoor pollutant levels  are
needed.
  Pollutants  within buildings are either
generated Indoors or  enter the building
envelope  from  outside. Examples of  in-
door-generated pollutants are nitrogen di-
oxide from gas-fired devices and formal-
dehyde from pressed-wood products. Ex-
amples of outdoor pollutants include sulfur
dioxide  and ozone. Some  pollutants are
not easy to classify as coming from either
indoor  or  outdoor sources, such  as car-
bon monoxide, which  can be generated
outdoors by automobiles and indoors by
burning tobacco  products  or residential
wood combustion.
  For  pollutants that  enter the building
envelope  from outside, the entry may be
either intentional or  unintentional. If the
exchange of outside air and hence outside
pollutants  is intentional, the process is
termed  ventilation. Examples of ventila-
tion include open windows, fresh- air vents,
and heating, ventilating, and air condition-
ing  (HVAC) systems.  When air enters a
building unintentionally,  the process is
called  infiltration. Likewise, the uninten-
tional venting of inside air to the exterior
is called exfiltration. Infiltration/exfiltration
occurs  due to defects in building  con-
struction. Cracks in exterior walls and leaks
around  windows and doors, for example,
allow outside air to enter and exit a build-
ing. The infiltration and exfiltration of air to
and from a building are driven primarily by
(1) the force of local winds on the building
exterior and (2) the  difference  between
inside and outside temperature.  The force
of local winds on a building, termed wind
pressure, is proportional to the square of
the wind speed and is controlled by both
building geometry and local  meteorologi-
cal factors. This wind pressure on the
building face induces a pressure gradient
across  it,  which results in  an air flow
through any openings  on the face; the
flow through an  opening is related to the
nature of the opening. Flow through such
openings may also be induced  by  differ-
ences between  inside and outside tem-
peratures that cause density differences
across the  building face. This process is
referred to  as the stack effect.
  The total exchange of air between inside
and outside for a building  during a given
time period, from both intentional and un-
intentional  sources,  is referred to as the
air exchange  rate of that building. This
.exchange rate-is-usually expressed-in the
number of  air changes  per  hour (ACH),
which represents the number of theoreti-
cal building volumes exchanged per hour.
   Indoor pollutants may undergo chemical
reactions and  loss by adsorption and ab-
sorption to  building furnishings. Pollutants
are also dispersed through the building as
a function of interroom flows. The disper-
sion of pollutants through the building is a
function of  many building and local envi-
ronment factors.  Of course, new pollutants
of  interest may be formed from  any
chemical reactions that occur.
  Attempts have been made to measure
indoor  pollutant levels  within buildings,
usually involving either personal exposure
monitors (PEMs) or complex tracer stud-
ies. However, the former method does not
readily lend itself to a detailed cause-and-
effect analysis, while the latter is difficult
and expensive to perform. Deducing how
individual  parameters  affect indoor air
concentration  levels, therefore, is often
difficult. To aid in such analysis, physically
relevant models have been developed to
simulate indoor  air  concentration levels.
These models  use mass balance concepts
to estimate pollutant dispersion  within
buildings based on first- order chemical
kinetics. Such mass balance models re-
quire a detailed description of the building
to be modeled, as well as many inputs not
commonly   available.  To  address  such
complications, simpler  statistical models
requiring less  detailed inputs  have also
been used  in attempts to estimate indoor
air pollution levels.
  This study, organized into two sections,
attempts to clarify some of the issues dis-
cussed above that affect indoor air quality.
In the first  section results from two mass
balance   models,  CONTAM87  and
CONTAM88, are investigated. Both mod-
els were developed  by the National Insti-
 tute of Standards and Technology (NIST),
 formerly the National Bureau of Standards
 (NBS). The  models were used here  to
 simulate  indoor exposure to ozone con-
 centrations in a hypothetical four-story of-
 fice building. The section seeks to answer
 some of the  following questions concern-
 ing indoor exposures:
    1. Do  modeled  indoor air concentra-
      tions indicate sufficient room-to- room
      variation in pollutant concentrations
      to justify the  resources  required  to
      use such sophisticated models?
    2. If significant variations are indicated,
      what factors influence them?
    3: What impacts do possible variations
      have on requirements for measuring
      pollutants within  indoor environ-
„ ...	ments?  	.,,_,-.		
    4. How practical is it to use such mass
      balance models?
   The second section  investigates the  is-
 sue of using  statistical  models to estimate
 indoor exposures, particularly the need to
 account  for  serial correlation  in such
 models.

 Mass Balance Simulation of
 Indoor Exposures
   To investigate the effects of certain pa-
 rameters on indoor air concentrations, two
 mass balance simulation models were  in-
 vestigated CONTAM87 and CONTAM88.
 All simulations were centered on a hypo-
 thetical four-story office building with out-
 door ozone as the pollutant of interest.
   CONTAM87 is an indoor air quality dis-
 persion model developed in  1988 by the
 National Institute of Standards and Tech-
 nology (NIST). The model solves a series
 of simultaneous nonlinear equations gov-
 erning the dispersion of pollutants  among
 a  network of well-mixed "zones." These
 well-mixed zones typically represent indi-
 vidual rooms, but may also represent other-
 separate entities, such  as  garages and
 HVAC systems. The zones are connected
 by air flow elements that define the ex-
 change  of pollutant-laden  air between
 zones. The user must define the interzonal
 flows, which  is not a  trivial task.  In this
 study, the interzonal floWs were estimated
 using "rules of thumb" proposed  by the
 American Society  of Heating, Refrigera-
 tion and Air Conditioning Engineers.
   CONTAM88 is a hybrid of the previous
 model and was developed in 1990 by the
 NIST.  CONTAM88 also represents a
 building as a collection of well-mixed zones
 connected by flow elements and solves a
 set of simultaneous equations to estimate
 pollutant dispersion. CONTAM88, how-

-------
ever,  uses a  mixture  of  analytical  and
empirical relationships for various building
components (doors, walls,  etc.) to solve
for the interzonal flows, based upon  out-
side meteorological conditions and physi-
cal  building characteristics.  These algo-
rithms were extracted from AIRMOVE, a
model developed by  NIST in 1972.
CONTAM88  includes  a  preprocessor
NBSAVIS, which is a user interface  that
allows one to "construct" a building from
its individual components.
  The building modeled was a four-story
office  building adequate for a total occu-
pancy of 131 persons.  The building con-
sisted of four  identical floors,  each  with
nine offices, two  restrooms, and two stair-
way entrances. Each floor measures 26
m x 18 m x 3 m  for a volume  of 1404
cubic  meters. On each floor, eight exterior
offices, two restrooms and two stairwells
open  onto an inner hallway. The inner
hallway surrounds an interior office. Each
office, except for the interior office located
in the center of each  floor, has  one or
more  exterior windows.  The interior office,
both stairways, and both restrooms have
no windows. Entrance  is gained  into the
building via two exterior  doors,  one in
each first-floor stairway.
  One of the  first questions to be  ad-
dressed was whether using such  detailed
indoor air quality  models is necessary.
Could a simpler, more manageable way
to calculate indoor exposure  be used?
Findings from this study indicated that for
the office  building  considered there was
substantial spatial  variation in  modeled
indoor ozone  concentrations. Modeled
concentrations showed an order-of-mag-
nitude variation from room to room, even
on the same floor. This suggests that in
some  instances simple  one-cell box mod-
els may not be adequate to simulate indoor
exposures for persons inside office build-
ings.   Study  results  revealed that
CONTAM88 predicted  maximum indoor
ozone concentrations 400% above those
predicted by the  one-cell box model.  The
study  also indicated that the one-cell  box
model underestimated mean building con-
centrations by  20% to 25% compared to
the more  detailed  mass balance model.
These results suggest that the complexity
of such models may be necessary to esti-
mate  exposures within buildings, even for
calculating average exposures.
  In addition to  the  above  findings,  we
also found that the building configuration
played a significant role in the  level of
indoor exposure. The simple act of open-
ing all interior office doors nearly doubled
the air exchange rate of the building and
raised maximum ozone concentration lev-
els by 30%. The opening of interior doors
also changed the magnitude and location
of the maximum concentration. Compari-
sons with the one-cell box model indicated
that both models predicted average con-
centrations proportional to changes in the
overall air exchange rate. CONTAM88 re-
sults, however,  indicated  that maximum
concentration values responded less pro-
portionally to changes in the air exchange
rate than did mean concentration values.
More work  is needed to examine  such
factors as the effect of temporal wind di-
rection variation on modeled  concentra-
tions. Future work is planned to address
such issues.
  Our findings suggest that persons inter-
ested in measuring  indoor pollutant levels
should carefully consider  the  placement
of monitors within the study building. Con-
ducting  simulations using physically rel-
evant models prior to designing the ex-
periment is  recommended. For example,
results from our study suggest a minimum
of five monitors  per floor to  accurately
characterize concentrations within the hy-
pothetical office building we modeled. The
study indicated that concentrations gener-
ally  decreased in magnitude toward the
building interior, but also depended  upon
wind direction and building configuration.

Statistical Modeling of Indoor
Exposures
  Correlating the temporal  and spatial
concentration variations within  microenvi-
ronments with the locations of people is a
very complex process. As a result, statisti-
cal methods of  estimating concentration
values for  use in human  exposure esti-
mates have  become more popular. The
number  of  variables involved and  their
associated uncertainties make determinis-
tic  models difficult  to use. SHAPE and
NEM, which include statistical methods
for characterizing concentration values for
different microenvironments, are examples
of human exposure models that combine
deterministic and statistical methodologies.
In SHAPE, concentrations are estimated
by sampling from a concentration distribu-
tion  for a  given  microenvironment in  a
Monte Carlo fashion. In model simulations
conducted thus far, however,  correlation
effects have been  ignored. The model
developers recognized that this tends to
underestimate maximum exposures, since
the concentration in a given microenviron-
ment is related to source terms and atmo-
spheric processes. The current sampling
procedures also tend to underestimate for
another  reason, referred  to as  the
"intracluster" correlation effect. Intracluster
correlation occurs for individuals in some
"dirty" microenvironment, such as a com-
mute that exposes  them to particularly
high carbon monoxide levels, because the
individuals tend to be exposed to these
high levels every commute. It has been
suggested that since Monte Carlo simula-
tions conducted thus far have  made no
provision for intracluster correlation effects,
they have underestimated the highest ex-
posures and  overestimated  the lowest
exposures.
  Recent investigations using Monte Carlo
simulations of exposure have suggested
that serial correlation may be too signifi-
cant to ignore. The purposes of  our study
were to quantify the  factors affecting se-
rial correlation in indoor microenvironments
and to investigate, in a very preliminary
way,  how personal exposure monitoring
(PEM) data can be used to infer the values
of variables needed  to estimate indoor •
concentrations.
  Indoor concentration values were simu-
lated using a mass-consistent box model
driven by a 48-hour sulfur-dioxide  data
set collected durihg the RAPS study. One
thousand 8-hour averages were gener-
ated  for three different scenarios.  The
Monte Carlo simulations underestimated
the highest concentrations by about 20%
and overestimated the  lower concentra-
tions  by about 30%.  The difference be-
tween using each hour's actual  air ex-
change rate versus  an average  air ex-
change rate for  each microenvironment
during the 48-hour period was slight. This
suggests that a model like NEM could be
updated  to include  the effects of serial
correlation, since a mass- consistent box
model could be included rather easily.
  We  also conclude  that personal expo-
sure monitoring data may be useful for
deriving order-of-magnitude estimates of
rate constants. More study is needed to
ascertain whether the positive correlations
and relative magnitudes of the time-aver-
aged cross-product terms would behave
as seen in our analysis if real-world data
were used. Our findings suggest that, due
to correlation effects, the use of PEM data
with corresponding fixed-site data would
underestimate  air exchange rates, under-
estimate emission rates, and overestimate
removal rates  where only one or at the
most two processes are in effect.
                                                                                 •&U. S. GOVERNMENT PRINTING OFFICE: 199 1/548-028/20204

-------
  John S. Irwln (also the EPA Project Officer,  see below), and WHHan B. Petersen,
    are with the Atmospheric Research and Exposure Assessment Laboratory, Re-
    search Triangle Park, NC27711; Ronald E. Stogner, Mourad Aissa and Azzedine
    Lansariare with Computer Sciences  Corporation, Research Triangle Park, NC
    27709.
  The complete report, entitled "Two Indoor Air Exposure Modeling Studies: CONTAM
    Modeling Results, and Serial Correlation Effects," (Order No. PB91-159 707/AS;
    Cost: $15.00, subject to change) will be available only from:
          National Technical Information Service
          5285 Port Royal Road
          Springfield, VA 22161
          Telephone: 703-487-4650
  The EPA Project Officer can be contacted at:
          Atmospheric Research and Exposure Assessment Laboratory
          U.S. Environmental Protection Agency
          Research Triangle Park, NC 27711
 United States
 Environmental Protection
 Agency
Center for Environmental
Research information   ..
Cincinnati, OH 45268
      BULK RATE
POSTAGE & FEES PAID
         EPA
   PERMIT No. G-35
Official Business
Penalty for Private Use $300
EPA/600/S3-91/013

-------