You are on page 1of 14

Journal of the Air & Waste Management Association

ISSN: 1096-2247 (Print) 2162-2906 (Online) Journal homepage: https://www.tandfonline.com/loi/uawm20

The Effect of Opening Windows on Air Change


Rates in Two Homes

Cynthia Howard-Reed , Lance A. Wallace & Wayne R. Ott

To cite this article: Cynthia Howard-Reed , Lance A. Wallace & Wayne R. Ott (2002) The Effect
of Opening Windows on Air Change Rates in Two Homes, Journal of the Air & Waste Management
Association, 52:2, 147-159, DOI: 10.1080/10473289.2002.10470775

To link to this article: https://doi.org/10.1080/10473289.2002.10470775

Published online: 27 Dec 2011.

Submit your article to this journal

Article views: 9515

View related articles

Citing articles: 40 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=uawm20
TECHNICAL PAPER Howard-Reed,
ISSN 1047-3289 J. Air & Waste Wallace, and
Manage. Assoc. Ott
52:147-159
Copyright 2002 Air & Waste Management Association

The Effect of Opening Windows on Air Change Rates in


Two Homes
Cynthia Howard-Reed and Lance A. Wallace
National Exposure Research Laboratory, U.S. Environmental Protection Agency, Reston, Virginia

Wayne R. Ott
Departments of Statistics and Civil and Environmental Engineering, Stanford University, Stanford, California

ABSTRACT house and 0.41 hr–1 (SD = 0.19 hr–1; n = 203) for the Vir-
More than 300 air change rate experiments were com- ginia house. Indoor/outdoor temperature differences ap-
pleted in two occupied residences: a two-story detached peared to be responsible for the variation at the Virginia
house in Redwood City, CA, and a three-story townhouse house of 0.15–0.85 hr–1 when windows were closed. Open-
in Reston, VA. A continuous monitor was used to mea- ing a single window increased the State 0 air change rate
sure the decay of SF6 tracer gas over periods of 1–18 hr. by an amount roughly proportional to the width of the
Each experiment first included a measurement of the air opening, reaching increments as high as 0.80 hr–1 in the
change rate with all exterior doors and windows closed California house and 1.3 hr–1 in the Virginia house. Mul-
(State 0), then a measurement with the single change from tiple window openings increased the air change rate by
State 0 conditions of opening one or more windows. The amounts ranging from 0.10 to 2.8 hr–1 in the California
overall average State 0 air change rate was 0.37 air changes house and from 0.49 to 1.7 hr–1 in the Virginia house.
per hour (hr–1) (SD = 0.10 hr–1; n = 112) for the California Compared with temperature differences and wind effects,
opening windows produced the greatest increase in the
air change rates measured in both homes. Results of this
IMPLICATIONS study indicate the importance of occupant window-open-
In estimating human exposure to airborne pollutants, it is
ing behavior on a home’s air change rate and the conse-
essential to understand how human behavior affects air
quent need to incorporate this factor when estimating
change rates in homes. We spend, on average, 66% of
the time in our own homes, and therefore much of our human exposure to indoor air pollutants.
inhalation exposure is due to residential indoor air con-
centrations of pollutants from indoor and outdoor sources. INTRODUCTION
Air change rates have a large effect on indoor concentra- The ability to estimate indoor and outdoor contributions
tions. An investigator trying to predict air change rates
to residential concentrations of air pollutants depends on
from meteorological conditions might incorrectly assume
that air change rates are at a minimum during relatively knowledge of a home’s air change rate. Air change rates
mild weather conditions (low wind speed and low out- have been measured in several probability-based U.S. stud-
door/indoor temperature differences) and a maximum ies, including more than 1360 homes in southern Califor-
during more extreme weather conditions. In contrast, it nia1-4 and more than 520 homes in Maryland and New
has often been observed that air change rates in occu-
York.5,6 The 5th percentile air change rates measured across
pied houses are at their highest, not lowest, when weather
conditions are more mild. Several previous investigators
all studies ranged from 0.13 to 0.76 hr–1; the median (50th
have speculated that this is due to increased window- percentile) air change rates ranged from 0.37 to 1.9 hr–1;
opening behavior under mild conditions. Our results quan- and the 95th percentile air change rates ranged from 1.1
titatively confirm that having a single window open, even to 6.1 hr–1. The explanations proposed for the variation in
during relatively mild meteorological conditions, can re-
these air change rates include meteorological conditions;7-11
sult in air change rates as high as, if not higher than, those
home structure differences;12-14 heating, ventilation, and
observed during extreme meteorological conditions. Thus,
it is important to account for occupant window-opening air-conditioning (HVAC) characteristics;15-19 and occupant
behavior when predicting residential exposures to con- behavior (e.g., window or door opening).20-22 Of these fac-
taminants, because opening windows significantly affects tors, the influence of occupant behavior on air change rate
air change rates and indoor concentrations. is one of the least understood, and yet it may account for a
significant portion of the variability. For example, based

Volume 52 February 2002 Journal of the Air & Waste Management Association 147
Howard-Reed, Wallace, and Ott

on the results of a survey of residential air change rates in A transom window located in the kitchen was occasion-
southern California, Wilson et al.23 suggested the measured ally tilted open several centimeters. Sliding glass doors (2
variability in air change rate was likely due to occupants m height) located in the nook and the dining room were
opening their windows. In another study, Kvisgaard and also opened during several experiments.
Collet21 estimated that 63% of the average air change rate The California house is located in Redwood City, CA,
in 16 Denmark homes was due to occupants opening win- which is ~48 km southwest of San Francisco, CA. This re-
dows and doors. gion is characterized by a narrow range of outdoor tem-
Opening windows is a relatively common occupant peratures year round. Based on 30-year meteorological data
activity as shown by the National Human Activity Pattern (1961–1990),25 ambient temperatures are relatively mild,
Survey (NHAPS) results.24 Of the 4723 people surveyed with monthly averages ranging from 9.2 ºC (January) to
evenly throughout the year, 42.5% of respondents opened 20.4 ºC (August). Hourly outdoor temperature and wind
at least one window in their home on the day of the sur- speed data were collected from the San Jose airport, ~39
vey. Of the 2006 respondents who opened windows, the km from the experimental house. During most air change
largest percentage (42%) opened one or two windows and experiments, temperature, wind speed, and wind direction
as many as 6% opened more than 10 windows. were not measured at the house. On occasion, outdoor and
Although the idea of opening windows to increase a indoor temperatures were measured at the California house
home’s air change rate is not new, there have been only on a 1-min basis with CO Databears (model T15, Langan
limited studies quantifying this effect. To better under- Products). Wind speeds were often noted on experimental
stand this effect, we measured air change rates in two oc- days and were categorized as calm, light, or gusty. Since
cupied houses in Redwood City, CA, and Reston, VA. We completing the experiments described herein, additional
used a real-time monitor to measure the decay rate of a wind data were collected at a height of 3.05 m in the back-
tracer gas (SF6), which provided air change rates over pe- yard at a distance of 4.6 m from the rear of the house using
riods of 1–18 hr. All experiments began by measuring air a computerized anemometer (WeatherPort).
change rates over a period of hours with all windows and The Virginia house is a three-story, four-bedroom, four-
doors closed (State 0), followed immediately by opening bathroom end-unit townhouse with a floor area of ~50 m2
one or more windows to different measured widths and per level (Figure 1b) and an approximate overall volume of
again measuring the air change rates after a new equilib- 400 m3. The townhouse is built on a slope such that there
rium was established. The difference in the measured air is a partial basement consisting of a utility room, pantry,
change rates associated with the two sets of conditions bathroom, and den with walkout patio. The middle level
was taken as the effect of the opened window. This ap- consists of a kitchen, dining room, living room, and bath-
proach minimized as much as possible the effects on air room. The top level contains four bedrooms (two of which
change rates of changes in meteorological conditions. have been converted to offices) and two bathrooms. The
home’s HVAC system uses 100% recirculated air and its
EXPERIMENTAL METHODOLOGY ductwork does not enter the attic, resulting in minimal
Experimental Houses duct leakage to the outdoors. Two “blower door” tests
Table 1 compares both study houses and locations. The [American Society for Testing and Materials (ASTM) Stan-
California detached house is a two-story, four-bedroom, dard E779]26 were recently performed to assess the air tight-
three-bathroom house with a floor area of 200 m2 and ness of the house. A blower door test consists of pressurizing
volume of ~510 m3. The living room, dining room, (or depressurizing) a building to a given pressure differ-
kitchen, and den are located on the first floor with con- ence and measuring the resulting airflow rates. The aver-
necting archways and no internal doors (Figure 1a). Also age closed house air change rate at 50 Pa was 14.2 hr–1 with
on the first floor is a bedroom and attached two-car ga- an estimated leakage area of 1120 cm2 at 4 Pa. The blower
rage. The living room has a “cathedral” ceiling extending door test also showed there to be negligible air exchange
to the second story. The second floor includes three bed- between the adjacent townhouses.
rooms and two bathrooms. The home’s HVAC system uses The windows of the Virginia house were also hori-
100% recirculated air; that is, there is no fresh air intake. zontal sliders and were similar in size to those in the Cali-
The windows in the California house were primarily fornia house. A majority of the experiments focused on
horizontal sliding glass panes. The experimental window the second floor dining-room window that was 1.4 m in
that was most often opened was located in the den and height. Other windows opened during experiments were
had a height of 1.4 m. Other windows opened during ex- located in the master bedroom, front bedroom, back of-
periments were located in the master bedroom, master fice, and recreation room. The heights of these windows
bathroom, upstairs bedroom, and downstairs bedroom. were also 1.4 m. A 2.0-m-tall sliding glass door located in
The heights of these windows ranged from 1.0 to 1.4 m. the living room was opened several times.

148 Journal of the Air & Waste Management Association Volume 52 February 2002
Howard-Reed, Wallace, and Ott

Table 1. Characteristics of the two houses studied in California and Virginia.

California House Virginia House

Location Redwood City, CA, midway between San Jose and San Reston, VA, 35 km northwest of Washington, DC
Francisco (42 km southwest of SFO airport) (13 km east of Dulles airport)
Weather Mild year-round, low winds Moderate temperatures, low winds
Type of House Detached, two-story End-unit three-story townhouse
Construction Materials Wood and stucco Cement block, stucco, and wood
Year of Construction 1986 1970
Floor Plan Four bedrooms, three bathrooms, attached garage, attic Four bedrooms, four bathrooms, partial basement, attic
Mixing Volume 510 m3 400 m3; ~135 m3 per level
Floor Area 200 m2 155 m2; ~52 m2 per level
HVAC Central a/c, gas heat Central a/c, gas heat
Window Types Horizontal sliding glass panes with screens Horizontal sliding glass panes with screens
Single Windows Opened Location Height (cm) Location Height (cm)
1 Den 137 Dining room 138
2 Nook sliding glass door 198 Recreation room 142
3 Dining-room sliding glass door 198 Office 144

The Virginia house is located in Reston, VA, which is Outdoor and indoor temperatures were measured on a
~35 km northwest of Washington, DC. In the District of 5-min basis using nephelometers (model 500, Climet In-
Columbia area, there is a greater temperature range be- struments), on a 1-min basis using CO Databears (model
tween winter and summer months than there is in Cali- T15, Langan Products), and occasionally with two house-
fornia. Monthly average temperatures (from 1962 to hold thermometers. All temperature monitors were col-
1993)25 ranged from –0.4 ºC in January to 24.4 ºC in July. located for several extended multiday periods to determine
Meteorological data were available on an hourly basis from individual bias, which was corrected to a single monitor.
nearby Dulles airport (~13 km from the Virginia house). The outdoor temperature measurements at the house were
significantly correlated with temperatures measured at
Dulles airport with a Spearman correlation coefficient of
0.95 (t = 188, n = 3995). During most air change rate ex-
periments, wind data were not collected; however, the
Virginia house was effectively shielded by tall trees (the
property backs onto a 0.12 km2 forested park) and nearby
townhouse rows, resulting in no noticeable wind a ma-
jority of the time. Subsequent to most of the window-
opening experiments, an ultrasonic anemometer
(Climatronics, Inc.) was installed on the roof (~1 m above
roof) of the Virginia house.

Tracer Gas Technique


In both houses, the air change rate was determined using
the tracer decay method. A detailed description of this
method is provided in ASTM Standard E741.27 Sulfur
hexafluoride (commercial product grade, 99.8% pure, Scott
Specialty Gases Inc.) was used as the tracer gas. In the
California house, a mass flow controller (model 5850E,
Emerson Electric Co., Brooks Instrument Division) was
used to regulate the total mass of SF6 injected. In the Vir-
ginia house, an amount of SF6 sufficient to fill the vol-
ume in the chamber between the two valves in the
Figure 1. Floor plans of (a) the California house and (b) the Virginia two-valve regulator was released in a single burst. For the
house. decay method, the exact mass of released SF6 does not

Volume 52 February 2002 Journal of the Air & Waste Management Association 149
Howard-Reed, Wallace, and Ott

need to be quantified because the calculated air change  C(t )  Qout (3)
ln =− t
rate depends only on the rate of change of tracer concen-  C(0)  V
trations. To avoid possible cylinder leakage, the tracer cyl-
inders were removed from the California house following where C(0) is the initial tracer concentration (ppm) and
each initial release. This was not found to be necessary at Q out
is the air change rate (hr–1). Using measured values
the Virginia house. V
of C over time, a plot of ln [C(t)/C(0)] versus time may be
Brüel & Kjaer (B&K) multigas photoacoustic analyz-
created. The negative slope of the best-fit straight line
ers (model 1302) were used to measure the SF6 concentra-
through these points is an estimate of the volumetric air-
tion decay in both houses. The B&K monitor uses an IR
flow rate divided by the house volume, which is typically
radiation beam to irradiate the gas sample with light at a
expressed as an air change rate with the units of air changes
frequency corresponding to an absorption frequency of
per hour (hr–1). Because the B&K monitors consistently
the gas. The gas absorbs some of the incident light en-
showed a positive reading under clean air conditions, and
ergy, the amount of energy depending on the concentra-
also in cases of nonzero residual levels of SF6 in the house,
tion of the gas. The absorbed light energy is reradiated at
it was important to subtract the appropriate background
a longer wavelength (heat), which results in periodic pres-
levels from all readings. This was done by averaging the
sure waves (also known as sound waves) that are mea-
B&K readings before the release of the tracer gas and also
sured by a microphone. A sample was collected and
after final equilibrium was reached and subtracting the
analyzed approximately every minute, providing “real-
mean of the two averages from all values.
time” measurements of the tracer concentration. The
monitor can detect SF6 at levels from ~0.1 to 80 ppm. The standard error was calculated for the negative
Two B&K monitors were operated side by side in the Cali- slopes of all best-fit lines based on the regression of eq 3. In
fornia house resulting in an average percent difference in general, the standard errors were relatively low. For example,
SF6 concentration measurements of 4.6%. In the Virginia the mean standard error averaged over 163 State 0 slopes
house, the B&K monitor was collocated with a gas chro- calculated for the Virginia house basement was 0.020 hr–1
matograph with electron capture detector (GC/ECD) (SD = 0.044 hr–1), or ~5% of the overall mean air change
(Valco Instrument Co., Inc.) to simultaneously measure rate of 0.41 hr–1. For open-window periods, the regression
SF6 concentrations. A bivariate analysis for 3 days of si- was often based on shorter (<1 hr) decay times and resulted
multaneous data collection showed the B&K monitor to in an average standard error of 0.074 hr–1 (SD = 0.11 hr–1),
average 97% (p < 0.001) of the SF6 concentrations mea- or ~7% of the mean open-window air change rate of 1.1
sured by the GC/ECD with an adjusted R2 value of 0.95. hr–1. Similar error values were observed for other levels in
Determination of a building’s air change rate is based the Virginia house and at the California house.
on the continuity equation.28-30 A general form of this Several tests were completed in both houses to assess
equation for each zone in a house may be written as the degree of tracer gas concentration uniformity among
and within the rooms. In the California house, separate

∑Q C − ∑Q C
dCiVi
= Qout,iCout − Qi,outCi + ji j ij i (1) tracer experiments were completed with simultaneous
dt
j j measurements in multiple locations. Results showed tracer
where Ci is the tracer concentration in zone i (ppm), Vi concentrations to be within ±2% on a single level and
is the volume of zone i (m3), Qout,i is the airflow rate within ±2% on different levels (ASTM Standard E74127
from outdoors to zone i (m3/hr), Cout is the tracer con- requires ±10% throughout the volume being tested). These
centration in outdoor air (ppm), Qi,out is the airflow rate results provide evidence that within 45 min of the tracer
from zone i to outdoors (m3/hr), Qji is the airflow rate release at closed conditions (State 0: no open windows or
from zone j to zone i (m3/hr), Cj is the tracer concentra- doors), the California house may be approximated by a
tion in zone j (ppm), and Qij is the airflow rate from single, well-mixed compartment. Thus, tracer concentra-
zone i to zone j (m3/hr). For the tracer concentration tions were usually measured in a single location for a given
decay technique using SF6, it is reasonable to assume experiment in the California house. However, the loca-
no contribution of SF6 mass from the outdoors (Cout = tion of the monitor was varied between experiments.
0). Also, each zone’s volume (Vi) may be assumed con- Unlike the California house, the Virginia house ex-
stant. For a single compartment (i = 1, j = 0), eq 1 may hibited a strong stack effect in the colder months result-
be further simplified to ing in a fairly uniform tracer gas concentration on each
floor with a gradient of concentration with floor level. As
dC Etheridge and Sandberg28 suggest, a reasonable character-
V = −QoutC (2)
dt ization of a multistory house is a single compartment per
Combining like terms and integrating eq 2 results in level. This division of the house was validated by separate

150 Journal of the Air & Waste Management Association Volume 52 February 2002
Howard-Reed, Wallace, and Ott

mixing experiments in which tracer concentrations were determinant. After a sufficiently long time, the portion
measured on different levels and at different locations on of the concentration associated with the smaller eigen-
the same level. In January, when the stack effect was at a values of eq 4 becomes negligible. At this point, as shown
peak, the following geometric means of the ratio of con- by Etheridge and Sandberg,28 the tracer decay rate in all
centrations for different levels with the central fan off compartments approaches a single value
were determined: Level 3/Level 1 = 1.5 (n = 13, GSD =
1.2), Level 3/Level 2 = 1.3 (n = 4, GSD = 1.1), and Level 2/ C(t) ≅ exp(max λkt) xmax (5)
Level 1 = 1.2 (n = 11, GSD = 1.1). In July, there appeared
to be improved mixing between the floors with the fol- where max λk is the maximum eigenvalue of the matrix
lowing geometric means: Level 3/Level 1 = 0.99, Level 3/ and xmax is the eigenvector associated with max λk.
Level 2 = 0.98, and Level 2/Level 1 = 1.0. It should be Thus, when plotting ln(Ci) versus time for each well-
noted that a potential reverse stack effect at the Virginia mixed compartment i, the concentration decay curve for
house during the summer months was counteracted by each compartment will become parallel, all with a slope of
the heat-generating equipment (pumps, computers, etc.) λk. The associated slope (max λk) is typically not the same
located on Level 1. During the summer months, tracer value as that of the whole-house air change rate. In fact,
concentrations were within 4% on any given level. Due Etheridge and Sandberg28 show the maximum eigenvalue
to the three-compartment nature of the three-level Vir- to be between the smallest air change rate and the largest
ginia house, in winter, the B&K monitor was retrofitted air change rate occurring in any zone. It appears that the
with an 18-m, 0.64-cm o.d. Teflon tubing such that mea- maximum eigenvalue is always as large as or larger than
surements could easily be made on different levels dur- the whole-house air change rate, and the overestimate is
ing the course of an experiment. Initial mixing of the tracer symmetrical whether a stack effect or reverse stack effect is
gas after release was normally enhanced through opera- occurring. Our three-compartment model indicates that,
tion of the home’s central fan. for a case in which the inflow of air at the lowest level is
Opening one or more windows also introduced a twice that at the highest level, the observed slope will be
multizone situation for both houses. An open window is 1.08 times the actual whole-house air change rate.
a source of “clean” outdoor air of different density than For experiments involving multiple compartments
indoor air. As a result, air entering the window rapidly (e.g., open-window conditions, State 0 in the Virginia
reduces the existing tracer concentration. This rapid re- house in winter), slopes for the decay curve were deter-
duction of concentration, however, was observed only in mined for periods of maximum eigenvalue domination.
the zone near the open window. For example, in the Cali- For the Virginia house, we defined the period of maxi-
fornia house, a monitor sampling in a nearby bedroom mum eigenvalue domination as the period after mixing
or upstairs did not record the rapid loss of tracer follow- when slopes for each zone were within 20% of each other.
ing the opening of the den window downstairs. A smoke Slopes that did not meet these criteria were not used in
tracer test in the Virginia house indicated that the flow of the data analysis.
air from the window was observable up to 6 m away from
the window, but not beyond. This new source of “clean” Approach
air eventually mixed with the remaining air volume of A “two-state” experimental approach was adopted and
the house; however, the tracer concentrations were dif- followed in both test houses. The first state, State 0, con-
ferent depending on the distance from the open window, sisted of all windows and external doors closed, with most
resulting in multiple compartments. internal doors open. For the California house experi-
When determining the air change rate for a multi- ments, State 0 also consisted of the door connecting the
compartment volume, a more complex method than attached garage and house being closed and the HVAC
that provided in eq 2 is needed. Sinden31 provides a de- system being off. Each house remained at State 0 long
tailed solution method for a multicompartment situa- enough to establish an associated well-mixed, “steady”
tion. This method results in the following tracer air change rate. State 0 time ranged from 3 to 10 hr in
concentration vector: the California house and from 30 min to 7 hr in the
n Virginia house.
C(t) = ∑α
k =1
k xkexp(λkt) (4) The second state incorporated a single change from
State 0 conditions; that is, a window was opened to a speci-
where n is number of compartments, αk is the solution fied width (1.3–63.5 cm). Again, measurement time was
constant dependent on initial conditions, xk is the eigen- sufficiently long to allow a valid measurement of the av-
vector associated with eigenvalue λk, and λk is the eigen- erage air change rate in this new state. This uninterrupted
value of matrix based on the solution of matrix dual-state approach minimized the effect of other cofactors.

Volume 52 February 2002 Journal of the Air & Waste Management Association 151
Howard-Reed, Wallace, and Ott

Additional experiments were completed in which mul-


tiple windows were opened on the same level as well as
on different levels.

RESULTS
Air change experiments were performed in the two occu-
pied residences between May 1995 and July 1999. A total
of 130 experiments were completed in the California
house and more than 200 experiments were completed
in the Virginia house. Between the two homes, experi-
ments were representative of different seasons, weather
conditions, and times of day. During this extended sam-
pling period, no major changes were made to either house. Figure 2. Comparison of air change rates with all windows and
doors closed (State 0) at the Virginia house and the California house.
State 0 Experiments For each class interval, the air change rate is equal to or greater than
(External Windows and Doors Closed) the value listed.

A total of 112 State 0 experiments were completed at the strongly bimodal, with high rates in the winter (Decem-
California house (Table 2). The measured State 0 air change ber 14, 1998, to March 21, 1999) (mean = 0.57 hr–1, SD =
rates ranged from 0.22 to 0.67 hr–1, with half of all mea- 0.14 hr–1, n = 95) and low rates in the summer (June 26 to
sured rates falling in a tight interval of 0.30–0.39 hr–1 (Fig- July 10, 1999) (mean = 0.15 hr–1, SD = 0.06 hr–1, n = 59).32
ure 2). The State 0 air change rates were primarily measured Because of its bimodal character, the complete distribu-
between 6:00 and 11:00 p.m. tion failed tests for normality and log-normality; how-
A total of 203 State 0 experiments, incorporating 393 ever, the individual winter and summer distributions were
measurements of air change rates for the three different not rejected for log-normality by the chi-square test (p =
levels, were completed at the Virginia house (see Table 0.47 and p = 0.37, respectively), and the winter distribu-
2). The measured State 0 air change rates were averaged tion was also not rejected for normality (p = 0.66). There-
across all three levels and ranged from 0.10 to 0.82 hr–1 fore, regressions were performed separately on the winter
(see Figure 2). The nighttime mean of 0.44 hr–1 (n = 121) and summer data. For the winter data (n = 74), the airport
was similar to the daytime mean of 0.39 hr–1 (n = 81). meteorological parameters accounted for 65% of the to-
The Virginia house was subjected to a wider range of tal variance. Of this total, 54% was associated with the
meteorological conditions than those experienced at the indoor/outdoor temperature difference, 9% with the wind
California house. speed, and ~1% each with the two orthogonal axes (north-
Multiple and stepwise regressions were performed on south and east-west). In summer, no meteorological vari-
the Virginia house air change rates to investigate the ef- ables significantly affected air change rates.
fects of temperature and wind speed and direction on the A similar multiple regression was not performed on
house at State 0 (closed windows). The meteorological pa- the California house. Instead, a series of nine experiments
rameters were obtained from Dulles airport, ~13 km north- were carried out in 2000 at the California house to look
west of Reston. The distribution of air change rates was very closely at the effect of wind on air change rate, using

Table 2. Seasonal averages of ventilation State 0 air change rates.

California House State 0 Virginia House State 0


Season Mean (hr–1) n SD (hr–1) SE (hr–1) Mean (hr–1) n SD (hr–1) SE (hr–1)

Winter 0.39 35 0.099 0.017 0.53 100 0.13 0.013


(Jan 1–March 31)
Spring 0.34 21 0.089 0.019 0.28 67 0.12 0.014
(April 1–June 30)
Summer 0.35 23 0.097 0.020 0.14 21 0.033 0.0073
(July 1–Sept 30)
Fall 0.38 33 0.11 0.019 0.55 15 0.16 0.040
(Oct 1–Dec 31)
Overall Average 0.37 112 0.10 0.0095 0.41 203 0.19 0.013

152 Journal of the Air & Waste Management Association Volume 52 February 2002
Howard-Reed, Wallace, and Ott

the backyard anemometer to Table 3. Single window/door-opening experiments at the California house.
measure wind speed and wind
direction. The approach was to Width of Area of Window/Door Date State 0 Air New Air Delta Air
measure air change rates at State Opening Opening Opened Change Rate Change Rate Change Rate
(cm) (cm2) (hr–1)a (hr–1)b (hr–1)c
0 on two consecutive days, one
windy and one nonwindy, to
2.5 343 Den Oct 20, 95 0.41 0.44 +0.03
see if a clear difference could be
3.8 521 Den Oct 26, 95 0.38 0.49 +0.11
noted. Temperature differences 5.1 699 Den Oct 13, 95 0.44 0.75 +0.31
from day to day were relatively 5.1 699 Den Nov 24, 95 0.50 0.79 +0.29
small. Three days included pe- 5.1 699 Den Jan 8, 96 0.49 0.78 +0.29
riods of several hours when the 7.6 1041 Den Oct 25, 95 0.39 0.64 +0.25
winds averaged approximately 10.2 1397 Den Oct 15, 95 0.46 0.76 +0.30
2–2.5 m/sec at the 3 m height, 10.2 1397 Den Nov 16, 95 0.44 0.79 +0.35
whereas for 6 days, the winds 10.2 1397 Den Feb 27, 96 0.48 0.55 +0.07
averaged less than 0.5 m/sec. No 10.2 1397 Den Mar 8, 96 0.25 0.60 +0.35
difference was seen in the mea- 10.2 1397 Den Apr 24, 96 0.24 0.59 +0.35
12.7 1740 Den Dec 22, 96 0.34 0.84 +0.50
sured air change rates on the
12.7 1740 Den Jan 15, 99 0.32 0.54 +0.22
windy versus nonwindy days.
15.2 2082 Den Sep 24, 95 0.29 0.46 +0.17
Thus, no effect of winds on air
22.9 3137 Den July 6, 96 0.36 0.73 +0.37
change rates was evident at the 27.9 3822 Den Aug 11, 96 0.30 0.56 +0.26
California house. 30.5 4179 Den Oct 11, 95 0.41 0.96 +0.55
33.0 4521 Den Aug 9, 96 0.32 0.61 +0.29
Single Window Opening 33.0 4521 Den Sep 29, 96 0.24 0.53 +0.29
Experiments 35.6 4877 Den Aug 20, 96 0.30 0.65 +0.35
A total of 30 experiments were 35.6 4877 Den Oct 2, 96 0.28 0.83 +0.55
completed with a single win- 38.1 5220 Den July 31, 96 0.35 0.82 +0.47
dow or door opening following 43.2 5918 Den Aug 21, 96 0.35 0.61 +0.26
a State 0 experiment at the Cali- 48.3 6617 Den Mar 17, 96 0.34 1.14 +0.80
50.8 6960 Den Oct 7, 96 0.29 0.82 +0.53
fornia house (Table 3). A typi-
74.9 14,830 Nook Oct 18, 96 0.28 0.74 +0.46
cal example of a plot of the
91.4 18,100 Nook July 18, 95 0.37 1.34 +0.97
concentration decay during the
91.4 18,100 LR door Aug 23, 95 0.38 1.05 +0.67
State 0 period and the subse- 91.4 18,100 Nook Oct 25, 98 0.33 1.08 +0.75
quent change in decay rate fol- 91.4 18,100 Nook Oct 27, 98 0.25 0.80 +0.55
lowing a window opening of
12.7 cm shows two distinct aState 0 air change rate is the steady-state air change rate with all external windows and doors closed; bNew air change rate is the steady-state
slopes representing each state air change rate following the opening of a window; cDelta air change rate is the difference between the new and State 0 air change rates.
(Figure 3). At the Virginia house,
34 single window-opening experiments were completed
(Table 4). At this house, during similar meteorological con-
ditions, a single window was opened on different floors re-
sulting in a varied effect on air change rate. For example, a
single window was opened 12.7 cm in the first-level recre-
ation room, resulting in an air change rate increase of 1.26
hr–1 (averaged across three levels). When a similar window
was opened 12.7 cm in the second-level dining room, the
air change rate increased by an average of 0.35 hr–1 (aver-
aged across three levels). A third-level office window opened
12.7 cm only increased the air change rate by 0.15 hr–1 (av-
eraged across three levels). This difference in the increase in
Figure 3. Representative “two-state” experiment at the California house,
air change rate might be attributed to the distribution of
in which State 0 was maintained between 14 and 17 hr, at which time the
pressure differences in the three levels of the house. Un- family-room window was opened 12.7 cm. The measured State 0 air
der stack conditions, more fresh air will enter a lower- change rate was 0.34 (±0.003 SE) hr–1 (n = 146, R2 = 0.992). The open-
level opening than a similar opening on the top floor. window air exchange rate was 0.84 (±0.003 SE) hr–1 (n = 116, R2 = 0.998).

Volume 52 February 2002 Journal of the Air & Waste Management Association 153
Howard-Reed, Wallace, and Ott

Table 4. Single window-opening experiments at the Virginia House. California house. Theory sug-
gests that if temperature differ-
Width of Area of Window Date State 0 Air New Air Delta Air ences are strong enough to
Opening Opening Opened Change Rate Change Rate Change Rate overwhelm effects due to wind
(cm) (cm2) (hr–1)a (hr–1)b (hr–1)c
and turbulence, then the
change in air change rate will
1.3 174 Dining room Mar 17, 97 0.59 0.71 +0.12
scale linearly with the product
2.5 355 Dining room Mar 31, 97 0.29 0.56 +0.27
2.5 355 Dining room Jan 4, 99 0.80 0.82 +0.02 of the window opening width
2.5 355 Dining room Jan 5, 99 0.76 0.98 +0.22 and the square root of the tem-
2.5 361 Recreation room Jan 6, 99 0.64 1.50 +0.86 perature difference;28 however,
2.5 361 Recreation room Jan 7, 99 0.62 1.50 +0.88 a plot of the delta air change
5.1 703 Dining room Feb 23, 97 0.54 0.81 +0.27 rate versus this compound
5.1 703 Dining room Mar 6, 97 0.44 0.67 +0.23 variable on the x axis failed to
5.1 703 Dining room Mar 15, 97 0.44 0.66 +0.22 improve the fit. It appears that
5.1 703 Dining room Dec 29, 98 0.52 0.89 +0.37
the relationship between air
5.1 703 Dining room Dec 30, 98 0.76 1.21 +0.45
change rate, temperature dif-
5.1 703 Dining room Jan 23, 99 0.26 0.52 +0.26
ference, wind and turbulence
5.1 729 Office Feb 16, 99 0.27 0.45 +0.18
7.6 1058 Dining room Mar 3, 97 0.33 0.73 +0.40 effects, neutral pressure height,
7.6 1084 Recreation room Feb 10, 99 0.60 1.90 +1.3 possible vertical temperature
10.2 1406 Dining room Mar 10, 97 0.57 1.07 +0.50 gradients, and other variables
12.7 1761 Dining room Dec 17, 98 0.45 0.96 +0.51 difficult to measure or hold
12.7 1761 Dining room Jan 27, 99 0.48 0.67 +0.19 constant prevent us from es-
12.7 1806 Recreation room Feb 6, 99 0.33 1.59 +1.26 tablishing a simple relation-
12.7 1826 Office Jan 12, 99 0.31 0.46 +0.15 ship between air change rate
15.2 2110 Dining room Mar 29, 97 0.17 0.62 +0.45
increases and the width of
15.2 2110 Dining room Apr 1, 97 0.64 1.03 +0.39
opening of a single window.
19.1 2639 Dining room Feb 3, 99 0.42 1.27 +0.85
20.3 2813 Dining room Mar 24, 97 0.54 1.14 +0.60
25.4 3516 Dining room Mar 28, 97 0.30 1.06 +0.76 Local Dilution Effect
25.4 3516 Dining room Dec 16, 98 0.42 1.0 +0.58 For several minutes immedi-
25.4 3516 Dining room Dec 21, 98 0.39 0.75 +0.36 ately following the opening of
25.4 3516 Dining room Dec 22, 98 0.67 1.35 +0.68 the California house den win-
30.5 4219 Dining room Mar 22, 97 0.41 1.15 +0.74 dow, an intermediate decay
40.6 5626 Dining room Apr 5, 97 0.37 1.06 +0.69 period with a steep slope was
50.8 7032 Dining room Apr 2, 97 0.38 1.27 +0.89 observed (Figure 5a). This
50.8 7032 Dining room Dec 15, 98 0.47 1.41 +0.94
rapid tracer decay, or “whoosh”,
50.8 7032 Dining room Dec 28, 98 0.87 1.66 +0.79
was measured during experi-
63.5 8794 Dining room Apr 7, 97 0.34 1.11 +0.77
ments in which the window
a b
State 0 air change rate is the steady-state air change rate with all external windows and doors closed; New air change rate is the steady-state was opened at least 2.5 cm
air change rate following the opening of a window; cDelta air change rate is the difference between the new and State 0 air change rates. and the B&K monitor was
sampling in the room con-
In both homes, a single window was selected to study taining the open window. As discussed earlier, this steeper
extensively the effect of opening a window to different slope reflects the influx of outdoor air to the immediate
widths ranging from 1.3 to 63.5 cm. In the California air volume, resulting in a significant dilution of the ex-
house, 25 experiments involved the opening of a win- isting tracer gas in only that zone. This concentration
dow located in the den (see Figure 1 for window loca- response was short-lived, typically lasting on the order
tion). In the Virginia house, a dining room window was of 10 min. As the concentration levels adjusted to this
selected for 28 controlled experiments. A monotonic in- change in conditions, the slope of the logarithm of con-
crease in air change rate with width of window opening centration became flatter and then stabilized at a value
was observed (Figure 4). A linear regression of the delta larger than the State 0 air change rate but less than the
air change rate and width of window opening resulted in value observed immediately after opening the window.
a slope of 0.012 ± 0.0016 hr–1/cm and intercept of 0.25 ± The values listed in Tables 3 and 4 represent air change
0.041 hr–1 for the Virginia house, and a slope of 0.0069 ± rates measured after the total house volume air had be-
0.0018 hr–1/cm and intercept of 0.19 ± 0.046 hr–1 for the come fully mixed.

154 Journal of the Air & Waste Management Association Volume 52 February 2002
Howard-Reed, Wallace, and Ott

Figure 4. Relationship between the increase in air exchange and window


opening width following the opening of a single den window in the
California house and a single dining-room window in the Virginia house.

Tracer decay results for the Virginia house also showed


this initial sharp decrease in SF6 concentration following
the opening of a window, but only on the floor where the
window was open. A representative tracer decay graph
(Figure 5b) shows several aspects common to the Virginia
house experiments. First, the slopes for each floor are par-
allel prior to the opening of the window (State 0), reflect-
ing equilibrium conditions. Second, the window was Figure 5. “Whoosh” effect observed at (a) the California house (family-
opened in the dining room at the same time the B&K room window opened 35.6 cm from time 24 to 31 hr) and (b) the Virginia
monitor was sampling in that room. As with the Califor- house (dining-room window opened 25.4 cm from time 14 to 16.8 hr).
nia house, a rapid clearing out (“whoosh” effect) was evi-
with every window and exterior door opened, the tracer
dent for a relatively short period of time. As seen in other
experiments, the lower and upper floors did not experi- gas concentration decreased to below the minimum
ence such a steep initial decline in tracer concentration detection limit within 2 min.
following a window opening in the dining room. Finally, Similar results were obtained for the multiple win-
the decay curve slopes were again parallel after a suffi- dow-opening experiments completed at the Virginia
cient mixing time had elapsed. house. The measurable sustained values ranged from 0.74
to 3.4 hr–1. The average increase in air change rate for all
Multiple Window Opening Experiments multiple window openings with an available start value
A more effective clearing out of the tracer gas was dem- was 1.25 hr–1 (SD = 0.49, n = 7).
onstrated when multiple windows were opened on dif-
ferent floors. In the California house, 29 experiments DISCUSSION
were completed with two or more window openings Historically, the estimation of residential air change rates
(Table 5). Transient period decay rates measured as high has focused on the influence of meteorological factors such
as 16 hr–1. Following the transient period, measurable as wind speed and indoor/outdoor temperature difference.
sustained air change rates (>40-min averaging time) As a result, we specifically looked at the relationship of
ranged from 0.10 to 2.8 hr–1. The average increase in air these factors and air change rates measured in the Cali-
change rate for all multiple window openings with an fornia and Virginia homes to compare their ability to in-
available starting value was 1.1 hr–1 (SD = 0.77, n = 21). crease air change rate with that of opening windows. In
The average increase in air change rate for opening dif- both experimental homes, the range of meteorologically
ferent combinations of two windows was 0.85 hr–1, influenced air change rates was within a single order of
which increased to 1.3 hr–1 when three or more win- magnitude, with a maximum value in the neighborhood
dows were opened. Often, the tracer gas concentrations of 0.8 hr–1. At the Virginia house, with all doors and win-
decayed to below detection limits within the first 40 dows closed, indoor/outdoor temperature differences were
min after opening the windows, not allowing the de- closely associated with measured air change rates. This
termination of a “readjusted” rate. For an experiment relationship is most easily seen by noting that the mean

Volume 52 February 2002 Journal of the Air & Waste Management Association 155
Howard-Reed, Wallace, and Ott

Table 5. Multiple window openings at the California house.

No. of Date No. of 1st No. of 2nd Monitor Initial Air Whoosh Regression Open Regression Delta Air
Open Floor Floor Location Change Decay Time Window Air Time Change Rate
–1 –1 –1 –1
Windows Windows Windows (hr ) (hr ) (min) Change (hr ) (min) (hr )

2 Apr 21, 96 1 1 1st floor 0.36 – – 0.84 110 0.48


2 June 27, 96 1 1 1st floor 0.33 – – 1.34 75 1.01
2 July 28, 96 1 1 1st floor 0.45 7.74 10 2.94 60 2.49
2 Aug 13, 96 1 1 1st floor – 4.79 20 1.62 65 –
2 Aug 31, 96 1 1 1st floor 0.49 10.7 5 1.22 50 0.73
2 Oct 9, 95 2 0 1st floor – – – 0.89 140 –
2 Apr 1, 96 2 0 1st floor 0.39 – – 0.87 65 0.48
2 Aug 30, 96 2 0 1st floor 0.36 5.19 10 1.71 100 1.35
2 Sep 1, 96 2 0 1st floor 0.35 4.67 10 0.69 350 0.34
2 Sep 3, 96 2 0 1st floor 0.36 4.26 10 0.73 270 0.37
2 Sep 7, 96 2 0 1st floor 0.36 3.37 20 0.77 120 0.41
3 June 27, 96 2 1 1st floor 1.34 – – 2.37 105 1.03
3 Mar 18, 96 1 2 1st floor – – – 2.53 70 –
3 Mar 28, 96 1 2 1st floor 0.33 – – 3.09 70 2.76
3 Mar 31, 96 1 2 1st floor 0.33 – – 1.76 170 1.43
3 Apr 8, 96 1 2 1st floor 0.32 – – 2.21 70 1.89
3 Apr 2, 96 3 0 1st floor 0.35 12 6 1.17 90 0.82
3 Apr 3, 96 3 0 1st floor 0.39 2.47 5 0.62 150 0.23
3 Nov 28, 98 3 0 1st floor 0.33 4.77 20 0.80 20 0.47
3 Nov 28, 98 3 0 2nd floor 0.37 4.95 20 0.47 20 0.10
3 Dec 13, 98 3 0 1st floor 0.35 6.15 21 2.07 100 1.72
4 Aug 18, 95 2 2 1st floor – 9.1 6 1.49 120 –
4 May 18, 96 2 2 1st floor 0.33 – – 1.66 210 1.33
4 May 28, 96 2 2 1st floor 0.35 5.06 10 2.31 60 1.96
6 May 19, 96 2 4 1st floor 0.29 6.67 40 – – –
7 Aug 29, 96 4 3 1st floor 0.33 15.8 12 – – –
7 Sep 14, 96 4 3 1st floor 0.37 7.5 35 – – –
7 Oct 8, 96 4 3 1st floor 0.31 9.02 32 2.11 40 1.80
8 Apr 20, 96 4 4 1st floor 0.29 11.2 30 – – –
8 May 14, 96 4 4 1st floor 0.25 11.9 30 – – –

wintertime air change rate was nearly 4 times the mean rate due to wind velocity and indoor/outdoor tempera-
summertime rate. When temperatures were nearly the ture differences is in the neighborhood of 0.7 hr –1,
same indoors and outdoors, the air change rate was ~0.1 whereas window opening can achieve extended peri-
hr–1, increasing to ~0.8 hr–1 when the temperature differ- ods with decay rates of ~2 hr–1. Thus, window opening
ence reached a maximum of nearly 30 ºC. Although wind in this area with fairly strong seasonal temperature
speed was significant in the wintertime regression on the variations can result in air change rates ~3 times those
Virginia house, it explained only 9% of the observed vari- due to the maximum indoor/outdoor temperature dif-
ance and was thus negligible compared with the tempera- ferences. At the California house, the dominance of
ture effect. In the case of the California house, the greatly window opening over meteorological effects is much
smaller range of indoor/outdoor temperature differences more pronounced, with window opening able to
made it difficult to determine a temperature effect, and a achieve air change rates almost 10 times higher than
wind speed effect was also not observable. the average meteorologically influenced (State 0) air
By adding the opening of a window to the already change rate. Thus, our results show that the air change
present meteorological influences of air change rate, rate in a single house is highly dependent on window-
we were able to further increase the air change rate by opening conditions and that a wide range of air change
an order of magnitude. Specifically, at the Virginia rates is achievable in a single house simply by adjust-
house, it appears that the maximum effect on air change ing the window positions.

156 Journal of the Air & Waste Management Association Volume 52 February 2002
Howard-Reed, Wallace, and Ott

This influence of window opening on air change rate They presented evidence that air change rates from at least
is an important consideration for indoor air quality mod- one study were significantly increased (1.22–1.34 hr–1)
els. In general, however, it is difficult to develop a quanti- when outdoor temperatures were between 19 and 24 ºC,
tative relationship between air change rate and the compared with the observed range of 0.35–0.98 hr–1 at
opening of one or more windows. According to the Ameri- temperatures outside this range. The marked increase was
can Society of Heating, Refrigerating and Air-Condition- attributed to the tendency of persons to open windows
ing Engineers (ASHRAE) Fundamentals Handbook,7 flow when the outdoor temperature is mild.
through an open window depends on several variables, Dick and Thomas33 studied air change rates in 28 oc-
including area of the window opening, number and loca- cupied houses during the winters of 1948 and 1949 in
tion of window openings, wind speed, indoor/outdoor England. Twenty homes were exposed to wind speeds on
temperature difference, and height of the neutral pres- the order of 3.8 m/sec and eight were in a more shielded
sure level (the point where the indoor pressure is equal to location with average wind speeds of 2.0 m/sec. The mean
the outdoor pressure). In addition, there is the presence air change rate under closed-window conditions in the
of turbulent eddies at the window opening. Although we homes with the higher mean wind speed was 1.51 hr–1,
found the increase in air change rate to be primarily de- compared with 0.88 hr–1 in the homes with the lower mean
pendent on window-opening area (width), many com- wind speed. The authors found that the effect of occu-
plex factors may need to be considered. pancy (i.e., window-opening behavior) was to increase
For example, the indoor/outdoor temperature dif- these mean rates by between 0.95 and 1.10 hr–1.
ference is also a function of the window opening width— Kvisgaard and Collet21 measured air change rates in
as the window opening width increases, the airflow 28 Denmark homes, of which 17 were naturally venti-
increases and replaces more of the indoor air with out- lated. Air change rates were measured for each house with
door air of a different temperature. In one case at the all windows and doors closed, resulting in a range of State
Virginia house, after opening the dining room window, 0 air change rates from <0.1 to ~0.4 hr–1 for the naturally
the temperature of that room declined from 23 to 9 ºC ventilated homes. When occupants were allowed to open
in 1 hr. At time t = 0 (measuring time from the opening windows and doors, the average air change rates increased
of the window), the temperature difference and, there- to 0.2–1.1 hr–1. Based on these measurements, the authors
fore, the flow is at a maximum; a monitor near the win- estimated that 63% of the total air change rate for these
dow will measure a rapid decline in concentration (the homes was due to occupants opening windows and doors.
“whoosh” observed in Figures 5a and 5b). At a later time, Iwashita and Akasaka20 concluded that 87% of the
tss, when steady-state conditions have again been ob- total air change rate was due to occupant behavior. The
tained, the same monitor will register a slower decrease. authors measured the air change rates in eight occupied
In the first case (t ~ 0), the observed decrease cannot be dwellings of a 16-story apartment complex with 52.2-m2
considered an air change rate (except possibly for the dwellings during the summer in Kagoshima City, Japan.
volume of air affected by the initial “whoosh”), but in The average State 0 air change rate, measured when the
the second case (t = tss), the final steady-state decrease is home was unoccupied with no windows open or HVAC
a true air change rate. It is possible that some of the operation, for seven of the eight dwellings was 1.9 hr–1.
higher air change rates reported in the literature are due The lumped average increase in air change rate due to
to this “whoosh” effect and are therefore overestimates. opening windows or operating the air conditioner was
~24 hr–1. These values are only upper bounds, however,
Window Opening Behavior because it is not possible to differentiate between loss to
Noted in Other Studies the outdoors and loss to hallways and adjacent apartments
In addition to being theoretically complex, there are few using a single tracer gas. Thus, it is not possible to di-
experimental data available to aid in understanding the rectly compare these findings with our results due to the
window-opening phenomenon. The relationship between difference in building type (i.e., high-rise apartment com-
air change rate and window opening width has not been plex vs. single family home).
quantified for real homes in the existing literature. How- Roulet and Scartezzini22 measured 8-hr average air
ever, several studies have observed a general increase in change rates in an occupied three-story detached house
air change rate when comparing measurements during (530 m3) in Lausanne, Switzerland, for a 12-day period.
periods a house was closed and periods windows were For the entire monitoring period, average air change rates
open. For example, Wilson et al.23 reviewed the results of were 0.37 hr–1 (SD = 0.083 hr–1, n = 13) when all of the
several California air change rate studies completed in windows were closed and increased to 0.55 hr–1 when one
occupied houses and concluded that window-opening or two windows were opened on the third level. The width
behavior accounted for much of the observed variability. of window opening was not specified, making it difficult

Volume 52 February 2002 Journal of the Air & Waste Management Association 157
Howard-Reed, Wallace, and Ott

to compare to results presented here. However, in the Vir- • The indoor/outdoor temperature difference was
ginia house, the increase in air change rate due to open- found to be the primary contributor to the Vir-
ing one upstairs window ranged from 0.15 to 0.18 hr–1, ginia house’s air change rate with all windows
consistent with the value of 0.18 hr–1 observed by Roulet closed (State 0).
and Scartezzini.22 • Opening a window by more than a few centime-
Traynor et al.34 also measured the increase in air ters often produces a rapid influx of air restricted
change rates when windows were opened in two unoccu- to a relatively small volume of the house for a
pied single-story homes in Albany, OR, that were con- transient period of a few minutes followed by a
nected by their garages. The State 0 air change rate for steady air change rate for the house as a whole.
both homes was ~0.2 hr–1. After opening all of the win- This whoosh effect (a temporary sharp drop in
dows (number not specified, but appears to be six win- tracer gas concentration near the window) is, of
dows per home based on floor plan drawing) in each home course, not the whole-house air change rate.
by 1 cm, the air change rate increased by ~0.3 hr–1. Inter- • A monotonic increase in air change rate with
estingly, opening one window by ~5 cm resulted in an width of window opening was documented. How-
average 0.3 hr–1 increase in air change in both the Califor- ever, the exact form of the relationship, for the
nia house and the Virginia house. complex conditions associated with the occupied
Window opening is not only an important determi- experimental houses, could not be determined.
nant of ventilation, but it also has a direct impact on con- • At the Virginia house, on all four occasions in
taminant concentration levels in homes. The increased the winter when a window was opened to the
infiltration of outdoor air associated with an open win- same width on different floors, the increase in
dow may reduce indoor contaminant concentrations as- whole house air change rate was greatest for the
sociated with indoor sources, as well as increase the case when a window was opened on the lowest
penetration of outdoor contaminants. Several studies have level. This finding suggests that the dependence
noted an increased correlation between measured indoor of air change rate on open windows is a function
and outdoor pollutant concentrations during periods with not only of the width of the opening but also
the windows open.35-37 Open windows may also affect the the elevation of the window.
distribution of pollutants in a building. For example, an • Opening multiple windows increased the air
occupant may open a single window to remove odors and change rate by as much as 2.8 hr–1 in the Califor-
smoke from a specific room (e.g., opening a kitchen win- nia house and 1.7 hr–1 in the Virginia house.
dow after a cooking fire). While this action may be effec- These increases are far greater than could be pro-
tive at removing a significant percentage of the pollutant duced in a closed house even by extreme winds
mass from the room with the open window, it does not and temperature differences.
rapidly remove pollutants from other affected areas in the • Future indoor air quality surveys should carefully
house. For example, after opening the dining-room win- take into account the window position of the
dow 25 cm in the Virginia house, the concentration in homes surveyed, because of the great importance
that room is reduced by 45% in 8 min, whereas it would of this main covariate over other factors and the
take ~45 min for the concentration upstairs to be reduced lack of emphasis given to it in past studies.
by the same amount. Because the heated smoky air from
cooking rises rapidly to the upstairs rooms, it may be ACKNOWLEDGMENTS
important to clear these rooms as well as the kitchen. Based The authors are grateful to Andrew Persily, Steve
on the experimental results of this study, a more effective Emmerich, Dan Greb, Steve Nabinger, and Stuart Dols of
method to evacuate pollutants from the entire house the National Institute of Standards and Technology (NIST)
would be to open multiple windows, at least one per level for supplying and maintaining a GC/ECD and SF6 tracer
of the home. gas for supplementary measurements during this study.
This research was partially funded by the Tobacco Related
CONCLUSIONS Disease Research Program of the State of California Grant
This study examined air change rates measured in two No. 6RT-0118 and by the U.S. Environmental Protection
different types of homes located on the east and west Agency (EPA) University Partnership Agreement No.
coasts of the United States and reached the following DW89931890 with Lawrence Berkeley National Labora-
conclusions: tory and Stanford University. EPA, through its Office of
• Of the covariates studied, opening windows and Research and Development, funded and managed part of
doors produced the greatest increase in air change the research described herein. It has been subjected to
rates measured in both homes. Agency review and approved for publication. Mention of

158 Journal of the Air & Waste Management Association Volume 52 February 2002
Howard-Reed, Wallace, and Ott

trade names or commercial products does not constitute 22. Roulet, C.; Scartezzini, J.L. Measurement of Air Change Rate in an
Inhabited Building with a Constant Tracer Gas Concentration Tech-
an endorsement or recommendation for use. nique; ASHRAE Trans. 1987, 93, 1371-1380.
23. Wilson, A.L.; Colome, S.D.; Tian, Y.; Becker, E.W.; Baker, P.E.; Behrens,
D.W.; Billick, I.H.; Garrison, C.A. California Residential Air Exchange
REFERENCES Rates and Residence Volumes; J. Expos. Anal. Environ. Epidemiol. 1996,
1. Özkaynak, H.; Xue, J.; Weker, R.; Butler, D.; Koutrakis, P.; Spengler, J. 6 (3), 311-326.
The Particle TEAM (PTEAM) Study: Analysis of the Data; Final Report; 24. Tsang, A.M.; Klepeis, N.E. National Human Activity Pattern Survey
Contract No. 68-02-4544; U.S. Environmental Protection Agency: (NHAPS); Report prepared for the U.S. Environmental Protection
Research Triangle Park, NC, 1996; Vol. III. Agency, Contract No. 68-W6-001, by Lockheed Martin: Las Vegas,
2. Colome, S.D.; Wilson, A.L.; Tian, Y. California Residential Indoor Air NV, 1996.
Quality Study. Vol. 2: Carbon Monoxide and Air Exchange Rate: A 25. WorldClimate Web site. http://www.worldclimate.com. Buttle and
Univariate & Multivariate Analysis; Integrated Environmental Services: Tuttle, Ltd.: 1997.
Irvine, CA, 1994. 26. Standard Test Method for Determining Air Leakage Rate by Fan Pressur-
3. Wallace, L.A.; Nelson, W.C.; Ziegenfus, R.; Pellizzari, E.D. The Los ization. ASTM E779-99; American Society for Testing and Materials:
Angeles TEAM Study: Personal Exposures, Indoor-Outdoor Air Con- West Conshohocken, PA, 1999.
centrations, and Breath Concentrations of 25 Volatile Organic Com- 27. Standard Test Method for Determining Air Change in a Single Zone by
pounds; J. Expos. Anal. Environ. Epidemiol. 1991, 1 (2), 37-72. Means of a Tracer Gas Dilution; ASTM E741-00; American Society for
4. Colome, S.D.; Wilson, A.L.; Becker, E.W.; Cunningham, S.J.; Baker, Testing and Materials: Philadelphia, PA, 2001.
P.E. Analysis of Factors Associated with Indoor Residential Nitrogen 28. Etheridge, D.; Sandberg, M. Building Ventilation: Theory and Measure-
Dioxide: Multivariate Regression Results. In Proceedings of the 4th In- ment; John Wiley & Sons: Chichester, United Kingdom, 1996.
ternational Conference on Indoor Air; Seifert, B., Esdorn, H., Fischer, M., 29. Breum, N.O. Diagnosis of Ventilation by Single-Tracer Gas Techniques;
Rüden, H., Wegher, J., Eds.; Institute for Water, Soil and Air Hygiene: Indoor Air 1993, Supp. 1, 3-28.
Berlin, 1987; Vol. I, pp 405-409. 30. Sherman, M.H. Tracer-Gas Techniques for Measuring Ventilation in
5. Wallace, L.A.; Manale, A.; Nelson, W.C. Preliminary Results of the a Single Zone; Bldg. Environ. 1990, 25 (4), 365-374.
Baltimore TEAM Study: II. Personal Air and Breath Measurements. In 31. Sinden, F.W. Multi-Chamber Theory of Air Infiltration; Building
Measurement of Toxic and Related Air Pollutants; EPA 600/9-88/015; Air Environ. 1978, 13, 21-28.
Pollution Control Association: Pittsburgh, PA, 1988; pp 149-154. 32. Wallace, L.A.; Emmerich, S.J.; Howard-Reed, C. Continuous Measure-
6. Sheldon, L.S.; Hartwell, T.D.; Cox, B.G.; Sickles, J.E., II; Pellizzari, E.D.; ments of Air Change Rates in an Occupied House for One Year: The
Smith, M.L.; Perritt, R.L.; Jones, S.M. An Investigation of Infiltration Effect of Temperature, Wind, Fans, and Windows; J. Expos. Anal.
and Indoor Air Quality; Final Report, Contract No. 736-CON-BCS-85; Environ. Epidemiol., recommended for publication.
New York State Energy Research and Development Authority: Albany, 33. Dick, J.B.; Thomas, B.A. Ventilation Research in Occupied Houses;
NY, 1989. J. Inst. Heat. Vent. Eng. 1951, 19, 306-326.
7. ASHRAE Handbook of Fundamentals; American Society of Heating, Re- 34. Traynor, G.W.; Apte, M.G.; Carruthers, A.R.; Dillworth, J.F.; Prill, R.J.;
frigerating and Air-Conditioning Engineers (ASHRAE): Atlanta, GA, 1997. Grimsrud, D.T.; Turk, B.H. The Effects of Infiltration and Insulation on
8. Murray, D.M.; Burmaster, D.E. Residential Air Exchange Rates in the the Source Strengths and Indoor Air Pollution from Combustion Space
United States: Empirical and Estimated Parametric Distributions by Heating Appliances; J. Air Pollut. Control Assoc. 1988, 38, 1011-1015.
Season and Climatic Region; Risk Anal. 1995, 15 (4), 459-465. 35. Abt, E.; Suh, H.H.; Allen, G.; Koutrakis, P. Characterization of Indoor
9. Chastain, J.P.; Colliver, D.G. Influence of Temperature Stratification Particle Sources: A Study Conducted in the Metropolitan Boston Area;
on Pressure Differences Resulting from the Infiltration Stack Effect; Environ. Health Perspect. 2000, 108 (1), 35-44.
ASHRAE Trans. 1989, 95 (1), 256-268. 36. Sarnat, J.A.; Koutrakis, P.; Suh, H.H. Assessing the Relationship between
10. Swami, M.V.; Chandra, S. Correlations for Pressure Distribution on Personal Particulate and Gaseous Exposures of Senior Citizens Living
Buildings and Calculation of Natural Ventilation Airflow; ASHRAE in Baltimore, MD; J. Air & Waste Manage. Assoc. 2000, 50, 1184-1198.
Trans. 1988, 94 (1), 243-266. 37. Long, C.M.; Suh, H.H.; Koutrakis, P. Characterization of Indoor Par-
11. Shaw, C.Y. Methods for Estimating Air Change Rates and Sizing Me- ticle Sources Using Continuous Mass and Size Monitors; J. Air & Waste
chanical Ventilation Systems for Houses; ASHRAE Trans. 1987, 93 Manage. Assoc. 2000, 50, 1236-1250.
(2), 2284-2302.
12. Parker, G.B.; McSorley, M.; Harris, J. The Northwest Residential Infil-
tration Survey: A Field Study of Ventilation in New Houses in the
Pacific Northwest. In Air Change Rate and Airtightness in Buildings;
ASTM STP 1067; Sherman, M.H., Ed.; American Society for Testing
and Materials: West Conshohocken, PA, 1990; pp 93-103.
13. Sherman, M.H. Analysis of Errors Associated with Passive Ventilation
Measurement Techniques; Bldg. Environ. 1989, 24 (2), 131-139.
14. Nazaroff, W.W.; Offermann, F.J.; Robb, A.W. Automated System for
Measuring Air-Exchange Rate and Radon Concentration in Houses; About the Authors
Health Phys. 1983, 45 (2), 525–537. Cynthia Howard-Reed (corresponding author) completed
15. Cummings, J.B.; Tooley, J.J., Jr. Infiltration and Pressure Differences this work as an environmental engineer with the Human
Induced by Forced Air Systems in Florida Residences; ASHRAE Trans.
1989, 95 (2), 551-560. Exposure Analysis Branch, National Exposure Research
16. Lambert, L.A.; Robison, D.H. Effects of Ducted Forced-Air Heating Laboratory (NERL) at the U.S. Environmental Protection
Systems on Residential Air Leakage and Heating Energy Use; ASHRAE
Agency (EPA). Dr. Howard-Reed currently works for the
Trans. 1989, 95 (2), 534-541.
17. Modera, M.P. Residential Duct System Leakage: Magnitude, Impacts, Building and Fire Research Laboratory at the National Insti-
and Potential for Reduction; ASHRAE Trans. 1989, 95 (2), 561-569. tute of Standards and Technology (NIST). Her mailing ad-
18. Parker, D.S. Evidence of Increased Levels of Space Heat Consumption
and Air Leakage Associated with Forced Air Heating Systems in Houses dress is NIST, 100 Bureau Dr., Stop 8633, Gaithersburg, MD
in the Pacific Northwest; ASHRAE Trans. 1989, 95 (2), 527-533. 20899-8633 (e-mail: chreed@nist.gov). Lance A. Wallace is
19. Robison, D.H.; Lambert, L.A. Field Investigation of Residential Infiltra- an environmental scientist with the Human Exposure Analy-
tion and Heating Duct Leakage; ASHRAE Trans. 1989, 95 (2), 542-543.
20. Iwashita, G.; Akasaka, H. The Effects of Human Behavior on Natural sis Branch, NERL, at EPA. Wayne R. Ott is an exposure
Ventilation Rate and Indoor Air Environment in Summer—A Field research scientist with joint appointments in the Depart-
Study in Southern Japan; Energy Bldgs. 1997, 25, 195-205. ment of Civil and Environmental Engineering and the De-
21. Kvisgaard, B.; Collet, P.F. The User’s Influence on Air Change. In Air
Change Rate and Airtightness in Buildings; ASTM STP 1067; Sherman, partment of Statistics at Stanford University.
M.H., Ed.; American Society for Testing and Materials: Philadelphia,
PA, 1990; pp 67-76.

Volume 52 February 2002 Journal of the Air & Waste Management Association 159

You might also like