Energy and Buildings 127 (2016) 206–216

Contents lists available at ScienceDirect

Energy and Buildings

journal homepage: www.elsevier.com/locate/enbuild

Occupant behavior regarding the manual control of windows in

residential buildings
Bongchan Jeong a , Jae-Weon Jeong b , J.S. Park b,∗
a
Department of Architectural Engineering, Graduate School of Hanyang University, South Korea
b
Department of Architectural Engineering, College of Engineering, Hanyang University, 222 Wangsimni-ro Seongdong-gu, 133-791, Seoul, 133-791, South
Korea

a r t i c l e i n f o a b s t r a c t

Article history: The objective of this study was to identify the relation between occupants’ daily activities and window
Received 19 January 2016 control and also to quantify the influence of environmental variables on occupants’ window opening and
Received in revised form 3 May 2016 closing behavior. A field measurement was carried out during non-heating period, from February to May
Accepted 31 May 2016
2015, and heating period December 2014 to February 2015 in twenty occupied housing units. Window
Available online 1 June 2016
state, indoor and outdoor environmental conditions were continuously monitored during the measured
periods. From the results of the field measurements, it was found that the control of window is strongly
Keywords:
related to occupants’ daily activities. Cooking, cleaning and getting fresh air accounted for 27%, 40%, and
Window control
Occupant behavior
33% of total openings, respectively. Window openings in relation to the activities occurred at the specific
Indoor air quality time during the day. The daily average opening frequency and opened hours were varied from 0.5 to 1.8
Thermal sensation numbers/day and from 0.4 to 0.6 h/day respectively depending on outdoor temperature. The proportion
Residential building of opened windows steeply changed when outdoor temperature exceeded 12.7 ◦ C. Window closing was
associated with the degree of drop in indoor temperature after windows were opened. The results of this
study can help developing a complete model inferring occupants’ window opening behavior in residential
buildings.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction
Residential buildings account for a large portion of energy con-
sumption and have significant energy saving potential. Due to this
energy saving potential, governments have enhanced regulations
related with the energy performance of the residential build-
ings. The residential buildings have become more tightly sealed
and insulated to minimize heat loss. And mechanical ventilation
systems with heat recovery units have been installed, because
infiltration or natural ventilation is often insufficient to meet the
minimum requirement for ventilation.
The mechanical ventilation systems are commonly equipped
with filters to remove particles from outdoor air and re-circulate
air. Park et al. [1] showed that mechanical ventilation systems with
filters can reduce indoor levels of airborne particles by 26% for sub-
micron particles and 65% for fine particles when compared with
naturally ventilated residential buildings. In terms of energy sav-
ing, several studies show that mechanical ventilation with heat
recovery can reduce energy consumption to provide a minimally
∗ Corresponding author.
E-mail address: junpark@hanyang.ac.kr (J.S. Park).
acceptable amount of outdoor air [2–4]. However, it was found from
Park and Kim’s study [5] that 68.3% (95 of 139) of occupants did not
use mechanical fans for ventilation at all, and the occupants fre-
quently used windows for ventilation more than the mechanical
fan. The results indicate that natural ventilation by opening win-
dows is still the predominant way occupants control the indoor
environment in residential buildings.
Opening windows is the most common and preferable nat-
ural ventilation method [6]. The two main parameters that are
influenced by opening windows are air change rate and indoor
temperature. Iwashita found that 87% of the total air change rate
was attributed to window and door opening [7]. Also Wallace
et al. found that window and door opening produced the greatest
increase in air change rates [8]. Opening windows is not only use-
ful for ventilation, but also provide a benefit for thermal comfort.
Brager et al. shows that there is a 1.5 ◦ C difference in the neutral
temperature felt between occupants with high and low degree of
window control [9]. Paciuk found that there is a significant relation
between the possibility of personal control and satisfaction [10].
These findings suggest that the understanding of the occupants’
interactions with windows is important to ensure a good perfor-
mance of naturally as well as mechanically ventilated residential
buildings. And the energy simulation based on the actual behavior

http://dx.doi.org/10.1016/j.enbuild.2016.05.097
0378-7788/© 2016 Elsevier B.V. All rights reserved.
 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216 207

Table 1
Characteristics of the sample housing units.

Sample ID Complex Floor area [m2 ] Floor levela Location Number of occupantsb Period of residence [year] Smokers Participated periodc

Of the building Within the building

a A 109 4/20 Border Inner 3 (M, F, m) 6.5 0 H/NH

b 109 8/20 Border Outer 4 (M, F, f, f) 6.8 0 H/NH
C 171 8/20 Border Outer 4 (M, F, f, f) 6.6 1 H/NH
d 109 16/24 Border Inner 4 (M, F, f, f) 7.0 0 H/NH
e 109 5/24 Border Inner 3 (M, F, f) 6.6 0 NH
f 129 14/26 Border Outer 4 (M, F, m, f) 7.0 0 H
g B 163 12/14 Border Outer 4 (M, F, m, f) 2.3 0 H/NH
h 163 4/15 Border Outer 4 (M, F, m, f) 2.0 0 H/NH
i 163 6/14 Border Inner 4 (M, F, m, f) 2.3 1 NH
j 163 10/15 Border Outer 4 (M, M, F, F) 2.0 0 NH
k 163 5/14 Border Inner 4 (M, F, m, f) 2.4 1 NH
l 136 2/19 Border Inner 4 (M, F, m, m) 2.8 0 H
m 145 4/19 Border Outer 4 (M, F, f, m) 3.0 0 H
n 145 13/19 Border Inner 4 (M, F, f, f) 1.2 0 H
o C 72 14/21 Border Inner 4 (M, F, f, f) 6.7 0 H/NH
p 80 7/23 Border Outer 4 (M, F, m, f) 9.5 1 H/NH
q 108 9/25 Center Outer 4 (M, F, f, f) 4.5 0 H/NH
r 80 6/20 Border Outer 4 (M, F, m, m) 1.4 0 H/NH
s 163 15/25 Center Outer 4 (M, F, m, f) 2.0 0 NH
t 79 19/25 Center Outer 4 (M, F, m, m) 16 0 H
a
Floor number/Total Number of floors.
b
M, male adult; F, female adult; m, male below 18 years; f, female below 18 years.
c
H/NH, both heating and non-heating period; NH, non-heating period only; H, heating period only.

Table 2
Measurement parameters and the specifications of devices used for measurement.

Parameter Device Accuracy Interval [minutes]

Outdoor Dry-bulb temperature Data Loggera (TR-72ui ±0.3 ◦ C 10

Relative humidity and RS-11, T&D) ±5% RH
Indoor Dry-bulb temperature Humidity, Temperature ±0.8 ◦ C 10
Relative humidity and CO2 monitor ±4% RH
CO2 concentration (MCH-383SD, Lutron) ±40 ppm
Particle number concentration Optical Particle Count (Aerotrack 9306, TSI) ±5% flow 60
Particle mass concentration Aerosol monitor (Dusttrak 8532, TSI) ±5% flow 60
Window opening Window opening status State Loggerb (UX90-001, Onset Computer) – Event
a
The sensors were radiation-shielded.
b
Binary window status was measured, opened or closed.

of occupants can provide reliable operational energy use which will
lead to more successful design of sustainable buildings.
Early studies used environmental parameters as input variables
to predict the windows opened. As early as 1951, Dick and Thomas
[11] found that outdoor temperature was the most important vari-
able affecting the number of windows opened. Raja et al. found a
change in opened windows started occurring at an outdoor temper-
ature of 15 ◦ C [12]. Few windows were opened when the outdoor
temperature was below 15 ◦ C, but increased steeply with an out-
door temperature above 15 ◦ C. The results of Nicol are consistent
with these findings [13]. The results of surveys for naturally ven-
tilated office buildings through different countries showed that
opened windows tends to increase significantly as the outdoor
temperature rises above 10 ◦ C. Recently, stochastic models for con-
trol of windows in office buildings have been proposed to reveal
affecting factors on window opening and closing behavior [14–18].
Window opening and closing behavior indicate change of window
state from closed to opened and from opened to closed respec-
tively. The main argument is that indoor and outdoor temperatures
are likely to be the key stimuli for opening windows, and building
characteristics may also affect the probability of windows being
opened.
In residential buildings, only a few studies have been made.
Levie et al. showed through the questionnaire survey that house-
hold characteristics, such as disposable income, house size, age and
gender of the occupants, and insulation level, are significantly asso-
ciated with window opening behavior in residential buildings [19].
Shi et al. showed that outdoor temperature is the most important
explanatory variable affecting the proportion of windows opened
in 8 residential apartments in China [20]. Andersen et al. studied
the probability of opening and closing windows in 15 Danish resi-
dential buildings and concluded that indoor CO2 concentration and
outdoor temperature are the two single most important variables
in determining the probability of opening and closing windows
respectively [21]. The result of Schweriker et al. showed that both
indoor and outdoor temperature are significant variables affecting
both opening and closing window probability [22]. Another impor-
tant finding was that specific calibration was required for buildings
equipped with an air-conditioning unit. These studies have consid-
ered environmental factors only to analyze occupants’ interaction
with windows control. The window opening behavior in residen-
tial buildings, however, can be strongly related to daily activities of
occupants, such as sleeping, cooking, and cleaning. The occupants’
activities vary during a day, and they can play an important role
in determining window opening and closing behavior. Thus win-
dow opening frequency, opening hours, and time of day are clearly
influenced not only by indoor and outdoor temperatures but also
the daily patterns of the occupants’ activities.
Based on the previous literature reviews with regard to win-
dow control, a field study was carried out in order to expand the
understanding of occupant interaction with window control in res-
idential buildings. The main objective of this study is to identify the
relation between daily activities and window control, and also to
quantify the influence of environmental variables on occupants’
208 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216

Fig. 1. Daily outdoor environmental conditions during measured period.

window opening and closing behavior. We monitored indoor and
outdoor environmental variables as well as conducted a question-
naire survey in 20 Korean housings for six months.
2. Methods
2.1. Samples
Twenty occupied housing units were selected from three multi-
family residential complexes located in a suburban area of Seoul.
The sample housing units were carefully recruited based on the
number of family members, floor area, location, floor number, and
period of residence as shown in Table 1. The three complexes con-
sist of several apartment buildings. The buildings are a typical flat
type with at least four single-story housing units on each floor. Dis-
trict hydronic radiant heating floors and electric air conditioning
devices are installed in all buildings.
Sample units a–f of Complex A has a mechanical supply fan
only for ventilation. Outdoor air is supplied from supply diffusers
mounted on ceilings through a single duct above the ceiling.
Exhaust vents are placed in kitchens and bathrooms. At complex
B, sample units g–n have mechanical supply fans and exhaust fans
with a heat recovery unit. The mechanically supplied air flow rate
is equal to the exhausted air flow rate. Sample units o–t of complex
C have local exhaust fans only, one in the kitchens and two in the
bathrooms. Outdoor and indoor air are exchanged mainly through
exhaust vents and uncontrolled leakage, or doors and windows
openings.
Each sample unit was occupied by a single-family, and most of
the families consisted of two adults and two children. The floor
areas and floor levels ranged from 72 to 171 m2 and 4th to 16th
 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216
Fig. 2. Daily indoor environmental conditions during measured period (the error bars indicate standard deviation).
respectively. Four units d, g, h and l had a smoker among family
members but they never smoked inside their homes. Each housing
unit has more than two external walls facing outdoors, and operable
windows are located in the external walls.
2.2. Measurements
The field measurement was conducted during a non-heating
period, from February to May 2015, and a heating period December
2014 to February 2015. Table 2summarized methods and devices
for measurements. The measurements consisted of three parts:
outdoor, indoor, and window position. Outdoor dry bulb temper-
ature and relative humidity were continuously recorded by a data
logger TR-72 and RS −11 T&D at one sample unit of each complex.
The indoor environment variables were measured in the living-
dining room of all sample units because cross ventilation by
209
window opening mainly occurs in the living-dining room. Indoor
dry bulb temperature, relative humidity, and CO2 concentration
were monitored continuously every 10 min by MCH-383SD Lutron.
Indoor particle numbers and PM10 mass concentration were hourly
measured during the day time by Aerotrack 9306, TSI and Dusttract
8532, TSI. The indoor environment monitoring devices were placed
on internal walls or furniture at a height of roughly 1.2 m above the
floor.
Window position was measured using state logger, UX90-001
Onset, to record when windows were opened or closed. The state
loggers were installed on two windows in the living room that were
frequently opened by respondents for cross ventilation purpose.
The windows monitored for this study were sliding type which is
the most-installed window type in residential buildings in Korea.
Additional climatic data that possibly affect occupants’ window
opening, such as solar radiation, wind speed, outdoor PM10 con-
210 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216
Fig. 3. Typical hourly average window opening and closing frequency for a day.
centration and rainfall, were collected from the closest station of
the Korea Meteorological Administration.
2.3. Subjective questionnaires
Twenty residents, one each from all sample units participated in
the questionnaire survey. The subjects were all female housewives
with no regular employment, thereby spending most of the time in
their homes. Their ages ranges from 35 to 57, with a mean of 44.7
years old. The survey information used for this study consisted of
three parts: (1) personal information, (2) general ventilation behav-
ior, (3) questionnaire on window opening activities.
The background survey obtained information about the sub-
jects’ personal data including gender, age and period of residence.
The ventilation behavior survey obtained information that possi-
bly affected the subjects’ window opening behavior. It included
mechanical ventilation system operation, window opening, and
air cleaner operation. The subjects were asked to fill out a ques-
tionnaire on activity sheets every time they opened windows. The
sheets had two questionnaires which asked about the purpose of
window opening and the thermal sensation. To evaluate thermal
sensation, ASHRAE seven-point scale which is cold (−3) to hot (+3)
with neutral (0) was used.
3. Results
3.1. Environmental variables
Fig. 1 shows the outdoor environment conditions during the
measured periods. Daily outdoor temperatures ranged between
−8.0–20.4 ◦ C and had no significant difference among the three
complexes. The daily outdoor temperature was below 10.0 ◦ C until
late February and rose steeply afterward. PM10 concentration and
wind speed ranged between 13–168 ␮g/m3 and 1.2 m/s–5.3 m/s
 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216
Fig. 4. Reasons to open windows at each hour of a day.
respectively. PM10 and wind speed had a negative correlation as
shown in Fig. 1(b) (p < 0.05).
Fig. 2 shows the changes of indoor air temperature and CO2 con-
centration during the measured periods. Indoor air temperature
ranged between 16.3 ◦ C to 26.7 ◦ C, which is higher than the outdoor
air temperature. The temperature was constant at a lower level dur-
ing the heating period and then increased along with the outdoor
temperature afterward. During the heating period, the indoor air
temperature was higher than 20.1 ◦ C on average, but there were
significant differences in among the sample units. The vertical bars
shown in Fig. 2(a) indicate standard deviation which means occu-
pants had different ranges of the indoor set-point temperature. The
difference of the indoor set-point temperature among the sample
units was higher than 6.0 ◦ C. The indoor temperature difference
among the sample units decreased after heating stopped. Fig. 2(b)
shows that the CO2 concentration had decreased throughout the
measured period. In the non-heating period, occupants opened
windows longer and more frequently and it resulted in lower CO2
concentration.
3.2. Daily activities and behaviors related to windows
In this section, a typical activity pattern related to windows
was analyzed. The window opening and closing frequency indi-
cates the change of the windows state, from closed to opened or
from opened to closed. When analyzing the window opening or
closing frequency, dataset was hourly averaged to infer the daily
behavior pattern during the measured period. Therefore the results
do not correspond to individuals, but rather the typical behavior
pattern of the occupants. Fig. 3(a)shows the hourly mean window
opening frequency. The earliest opening was observed at 5 am in
the non-heating period which is an hour earlier than in the heat-
ing period. The maximum frequency of window opening occurs at
10 am and 8 am in each period, respectively. This indicates that
the occupants’ behavior pattern shifted to earlier in the morning
in the non-heating period. This might be due to the early morn-
ing sunshine and higher outdoor temperature in the non-heating
period.
The peak change of windows from being opened to closed in
each period was at 11 and 9 am respectively, which is an hour later
than the maximum frequency of window opening in each period.
This shows that either windows were closed within an hour after
211
being opened, or windows were closed the day after they were
opened. In the non-heating period, more frequent window closings
in the evening were observed. Therefore windows were opened
earlier and closed late in the non-heating period which indicates
windows were opened longer. These results reveal a clear distinc-
tion of window behavior pattern compared to occupants in office
buildings. Window opening and closing patterns in office build-
ings are related to the presence of occupants [15]. The change of
the state of windows mainly takes place at 9 am when occupants
arrive at the workplaces. By comparison, window position changes
in residential buildings are more distributed and mainly occurred
between 8 am and 11 am. This suggests that the window behavior
pattern is time dependent but the frequency varies during the day
in relation to the period.
Fig. 4 provides the purpose of window openings according to
responds by the occupants whenever they opened windows. As
this survey was manually checked by the occupants, only 25% of the
total openings were recorded. The reasons to open windows were
divided into three main categories: cooking, cleaning and getting
fresh air. Cooking accounted for 27% of total openings and partic-
ularly occurred in the early morning and in the evening. Cleaning
accounted for 40% and takes place mainly around 10 am. This obser-
vation explains the time of the window openings during a day. To
get fresh air is comparatively widely distributed throughout the
day. The result indicates that activities are dominant window open-
ing stimuli, which explains 67% of total openings. Therefore the
occupants’ activity schedule is required to understand the window
opening or closing behavior in residential buildings.
Fig. 5(a)shows occupants’ daily average opening frequency (1 ◦ C
bin) distribution in relation to the outdoor temperature and stan-
dard deviation of occupants. The frequency varied on the average
between 0.5 and 1.8 numbers/day depending on the outdoor tem-
perature. The opening frequency was consistent at lower outdoor
temperatures and started to increase around 13 ◦ C. Fig. 5(b) shows
daily average window opened hours in relation to the outdoor tem-
perature. The average for the whole period was 1.8 h/day varying
between 0.4–6.0 h/day depending on the outdoor temperature. In
common with the result of Fig. 5(a), a steep increasing was observed
around the outdoor temperature of 10 ◦ C. These results of Fig. 5(a)
and (b) correlate to previous studies. Herkel [15] showed that the
lowest frequency and percentage of open windows were found in
winter and that the highest frequency in opening windows was
212 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216
Fig. 5. Daily average window opening frequency and opening hours in relation to outdoor temperature.
observed in spring and autumn. He also showed that the propor-
tion of windows opened increases strongly when reaching a certain
temperature. Findings by Nicol [23], through the surveys of the use
of simple controls of naturally ventilated buildings through dif-
ferent countries, showed that opened window tends to increase
significantly as the outdoor temperature rises above 10 ◦ C.
3.3. Window opening
From the previous sections, it was found that occupant behavior
patterns related with changing status of windows was time depen-
dent due to activities. In this section, the proportion of windows
opened was analyzed to evaluate what factors influence occupants
to keep windows opened. Fig. 6 shows the hourly average propor-
tion of windows opened during the day. The proportion of windows
opened indicates the number of data that windows are opened
divided by the total number of data. For example, the data point
at 9:00 at heating period indicates that the number of data that
windows were opened divided by total number of data from 08:30
to 09:20 during heating period. The proportion was observed much
higher in then non-heating period. The increase of the proportion
in the evening was much higher than in the morning. It indicates
that windows were opened longer in the non-heating period.
The influence of outdoor and indoor environmental variables on
occupants’ proportion of windows opened was analyzed using the
regressions as shown in Table 3. B is a regression coefficient used
for linear equation. ␤ is a standardized coefficient that refers to
the standardized level of influence on proportion of opened win-
dows. The outdoor air temperature has the strongest influence on
occupants’ opened windows (0.738). The second influential vari-
able is relative humidity of indoor air (−0.322). Negative influence
indicates that the less the relative humidity, the more windows
were opened. This can be explained by the time of day. The indoor
temperature varies mainly depending on the outdoor temperature
during the day. Windows are opened mostly during the daytime
when the indoor temperature is higher. Relative humidity has a
negative correlation with temperature due to change of amount
of saturated water vapor. This result shows that indoor relative
 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216 213

Fig. 6. Hourly proportion of opened windows in each period.

Table 3 cate that CO2 concentration was not a main stimulus, it was rather
Multiple regression analysis for window opening behavior with environmental
consequentially influenced by window opening.
factors.
The analysis on window opening frequency and opened hours
B ␤ Sig. shows that there was a steep change when the outdoor temper-
Outdoor environment ature is reached at a certain value. The proportion of windows
Temperature (◦ C) 0.014 0.738 0.00** opened as a function of outdoor temperature is modelled using the
Relative humidity (%) 0.001 0.107 0.00** change-point model. This method is capable of breaking data sets
Solar radiation (MJ/m2 ) 0.001 0.006 0.61
into segments and of inferring appropriate statistical models for
Wind speed −0.008 −0.084 0.00**
PM10 (␮g/m3 ) 0.000 0.031 0.00** each segment. Therefore it is suitable to detect whether change has
Rainfall (mm) 0.005 0.016 0.04* taken place. The functional form for the change-point regression
Indoor environment model is
Temperature (◦ C) 0.005 0.043 0.00**  +  +
Relative humidity (%) −0.007 −0.322 0.00** Pw open = C + CSH Tcp − Toa + CSNH Toa − Tcp (1)
CO2 Concentration (ppm) −0.000 −0.093 0.00**
PM10 (␮g/m3 ) 0.0000 −0.018 0.23 where C is constant term, CSH is the slope at heating, CSNH is the
*
p < 0.05. slope at non-heating, Tcp is the change-point temperature. P is a
**
p < 0.01. dependent variable for proportion of windows opened and Toa is
outdoor air temperature as an independent variable. The + nota-
Table 4 tion indicates that parenthetic term sets to be zero if a value is not
Logistic regression analysis for window closing probability with environmental positive. The proportion of windows opened and outdoor tempera-
factors. ture for 0.5 ◦ C bin are used for dependent and independent variables
Heating Non-heating
respectively to verify the effect of outdoor temperature on window
opening behavior. Equation obtained from change-point model is:
B Wald Sig. B Wald Sig.

Outdoor environment Pw open = 0.0811 − 0.0031(12.7 − Toa )+ + 0.0189(Toa − 12.7)+ (2)

Temperature (◦ C) −0.009 1.460 0.23 −0.036 21.736 0.00**
Relative humidity (%) 0.012 32.660 0.00** 0.000 0.004 0.95 The Eq. (2) explains that the relationship between opened pro-
Indoor environment portion and outdoor temperature. It shows that there is change
Temperature (◦ C) −0.054 14.063 0.00** −0.264 158.414 0.00** in the proportion of opened windows when the outdoor temper-
Temperature drop (◦ C) −0.072 8.912 0.00** 0.014 0.388 0.53 ature reaches the change-point temperature of 12.7 ◦ C. The first
Relative humidity (%) 0.001 0.027 0.87 0.006 1.117 0.29
CO2 Concentration (ppm) 0.000 6.059 0.01* −0.001 39.991 0.00**
slope and second slope, 0.0031 and 0.0189 respectively, indicate
that occupants open windows 6 times longer after the change-point
*
p < 0.05.
**
temperature.
p < 0.01.
Fig. 7 shows that the observed data and fitted curve using
change-point model (R2 = 0.98) compared with logistic regression
(R2 = 0.96). The logistic regression model was developed in previ-
humidity has a negative influence on opened proportion. Indoor ous studies [14], and the probability distribution is expressed by
temperature has less influence (0.043), because the range of the using logit function. The goodness of fit, R2 , value explains that
indoor temperature of each sample unit was significantly different the outdoor temperature can stand alone to infer the proportion
during the heating period as shown in Fig. 2(a). Outdoor relative of windows opened. The logistic regression analysis from various
humidity and PM10 concentration were found to have a positive countries resulted by recent studies are presented in Fig. 8. Nicol
influence but it is far lower than the outdoor temperature (0.107 investigated the occupant behavior of 25 office buildings in the UK
and 0.031). Wind speed also had a low negative influence. Indoor and Europe in 2001 [13]. The result showed that different propor-
CO2 concentration was found to be negatively linked. This may indi- tion of windows opened from the three countries was monitored.
214 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216

Fig. 7. Proportion of opened windows as a function of outdoor temperature with logistic regression and change-point model curve (0.5 ◦ C bin).

Fig. 8. Comparison of the curve obtained using change-point model with recent studies.

UK office workers generally tended to open windows more fre-
quently than other European countries. Another investigation of
window opening in UK office buildings by Rijal et al. [14] resulted
in a fewer opened proportion than Nicol’s investigation. The result
of this study showed a relatively lower opened proportion when
compared to the result of office buildings in other countries.
3.4. Window closing
In this section, the time when windows were closed was ana-
lyzed. We supposed that closing a window is such an action to quash
environmental discomforting stimuli. Logistic regression was used
to evaluate the influence of outdoor and indoor environmental
variables on the probability of changing the window status, from
opened to closed. Table 4 shows the result of the analysis. B is
a regression coefficient used for inferring probability of closing a
window. The temperature drop means the degree of the indoor
temperature dropped after the windows opened. However, it is
insufficient to evaluate the influence of each variable using B only,
because the variables have different scale. For example, outdoor
humidity and CO2 concentration have a range of 15–100% and
205–3293 ppm each. Thus the influence of each variable was tested
using Wald tests. In the heating period, outdoor relative humid-
ity, indoor temperature, and the temperature drop had significant
influence. In the non-heating period, indoor temperature, outdoor
temperature, and CO2 concentration were relevant. Although there
were slight differences in significance at each period, it was found
that indoor temperature is the most important variable to deter-
mine closing windows in both periods.
 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216
Fig. 9. Temperature drops when windows were closed.
Fig. 9(a) and (b) shows different thermal stimuli at each period
when closing windows. The indoor temperature drop indicates
the temperature difference between opening and closing windows.
Although each sample unit had different indoor set-point temper-
ature in the heating period, 90.6% of opened windows were closed
before the indoor temperature drop was reached at 4 ◦ C. In the
non-heating period, as shown in Fig. 9(b), the indoor temperature
drops show a good linear relationship with the outdoor temper-
ature. 84.6% of the data are included within ±2 ◦ C band from the
central regression line.
4. Discussion
The results of this study show occupant behavior related to
window opening and closing in residential buildings. Occupants
opened their windows when desiring fresh air or when doing activ-
ities. Window openings in relation to the activities occurred at the
specific time during the day. As a result, occupant behavior pattern
215
of changing a window status was time dependent due to the activ-
ities. It also indicates that the operation of a window is dependent
on the time of the day more than the physical variables only [16].
Occupants in un-air-conditioned space open windows for two
main reasons: to improve indoor air quality or to bring a cooling
effect by dropping the indoor temperature and stimulating indoor
air movement. During the heating period, it is hard to believe that
occupants open windows to bring a cooling effect. Therefore the
main reason for this period would be to improve indoor air quality.
Since occupants starts to expect a cooling effect, the window open-
ing pattern would differ as to achieve both IAQ and cooling effect.
To confirm this assumption, the change-point regression model
was used to detect behavioral change. As a result, Fig. 7 shows that
the second slope becomes larger than the first slope when outdoor
temperature exceeds the change-point 12.7 ◦ C.
The results obtained from this study show a lower rate of opened
windows than the results from other studies of office buildings as
presented in Fig. 8. In the daytime, the housewives were at home
216 B. Jeong et al. / Energy and Buildings 127 (2016) 206–216
alone in most cases, which means that fewer people had access to
the windows. Besides the subjects in their own space might be more
tolerant and satisfied with their indoor environment compared to
workers in an office.
A previous study suggested outdoor temperature as the most
influential factor on the proportion of opened windows [20]. Out-
door temperature was used as explanatory variable to predict the
time of closing windows [22]. However Table 4 shows that indoor
temperature and indoor temperature drop are better to be used for
closing windows than outdoor temperature during heating period.
It is not clear whether the outdoor temperature has direct influ-
ence on control of windows. But it is obvious that occupants are
affected by indoor temperature which is mixed with outdoor air
supplied through openings rather than outdoor air itself. A previ-
ous study showed that use of two strongly correlated variables on
a regression model may reduce quality of the model as one vari-
able dampens another variable’s regression coefficient [17]. Thus
further development is necessary to improve the validity of the
result.
The results of this study form a framework of general behavior
patterns in residential buildings which is important for predic-
tion of reliable operational energy and thermal conditions [24].
However, the measurements were conducted during heating and
non-heating period only. When outdoor temperature exceeds the
acceptable indoor temperature, it becomes hard to expect window
opening to bring a cooling effect. Accordingly, to verify the seasonal
effect, further measurements in other seasons, especially summer,
has been carried out to develop a complete model. A stochastic
modelling approach is now in progress and will be reported on
later.
5. Conclusion
This study monitored the occupant behavior patterns with
respect to window control in twenty occupied housing units during
heating and non-heating periods. There was clear a distinction of
window behavior pattern compared to occupants in office build-
ings. The frequencies of window openings were time dependent
due to the activities occurring during the day and varied according
to the period of the year due to environmental variables. The steep
increase of the proportion of opened windows was observed when
the outdoor temperature exceeded 12.7 ◦ C, indicating a change in
the occupant behavior presumably to bring a cooling effect. Win-
dow closing behaviors were explained by thermal stimuli. When
windows were opened in the heating period, 90.6% of the openings
were closed before the indoor temperature drop reached 4 ◦ C. In the
non-heating period, the temperature drop band of the occupants
was 4 ◦ C as a function of outdoor temperature. The data provided
from this study can help developing a complete model inferring
occupants’ window opening behavior in residential buildings.
Acknowledgements
This work was supported by National Research Foundation of
Korea (NRF) grant funded by the Korean government (MEST) (NRF-
2013R1A2A2A05005131).
References
[1] J.S. Park, N.-Y. Jee, J.-W. Jeong, Effects of types of ventilation system on indoor
particle concentrations in residential buildings, Indoor Air 24 (2014) 629–638.
[2] I.S. Walker, M.H. Sherman, Energy implications of meeting ASHRAE standard
62.2, ASHRAE Trans. 114 (2008) 505–516.
[3] D. Hekmat, H.E. Feustel, M.P. Modera, Impacts of ventilation strategies on
energy consumption and indoor air quality in single-family residences,
Energy Build. 9 (1986) 239–251.
[4] J.A. Leech, M. Raizenne, J. Gusdorf, Health in occupants of energy efficient new
homes, Indoor Air 14 (2004) 169–173.
[5] J.S. Park, H.J. Kim, A field study of occupant behavior and energy consumption
in apartments with mechanical ventilation, Energy Build. 50 (2012) 19–25.
[6] S. Barlow, D. Fiala, Occupant comfort in UK offices – how adaptive comfort
theories might influence future low energy office refurbishment strategies,
Energy Build. 39 (2007) 837–846.
[7] G. Iwashita, H. Akasaka, The effects of human behavior on natural ventilation
rate and indoor air environment in summer – a field study in Southern Japan,
Energy Build. 25 (1997) 195–205.
[8] C. Howard-Reed, L.A. Wallace, The effect of opening windows on air change
rates in two homes, J. Air Waste Manage. Assoc. 52 (2002) 174–185.
[9] G.S. Brager, G. Paliage, R. de Dear, Operable windows, personal control and
occupant comfort, ASHRAE Trans. 110 (2) (2004) 17–35.
[10] M. Paciuk, The role of personal control of the environment in thermal comfort
and satisfaction at the workplace, PhD thesis, University of
Wisconsin-Milwaukee, 1989.
[11] J.B. Dick, D.A. Thomas, Ventilation research in occupied houses, J. Inst. Heat.
Vent. Eng. 19 (1951) 279–305.
[12] I.A. Raja, J.F. Nicol, K.J. McCartney, M.A. Humphreys, Thermal comfort: use of
controls in naturally ventilated buildings, Energy Build. 33 (2001) 235–244.
[13] J.F. Nicol, Characterising occupant behaviour in buildings: towards a
stochastic model of occupant use of windows, lights, blinds, heaters and fans,
Seventh International IBPSA Conference (2001) 1073–1078.
[14] H.B. Rijal, P. Tuohy, M.A. Humphreys, J.F. Nicol, A. Samuel, J. Clarke, Using
results from field surveys to predict the effect of open windows on thermal
comfort and energy use in buildings, Energy Build. 39 (2007) 823–836.
[15] S. Herkel, U. Knapp, J. Pfafferott, Towards a model of user behaviour regarding
the manual control of windows in office buildings, Build. Environ. 43 (2008)
588–600.
[16] G.Y. Yun, K. Steemers, Time-dependent occupant behaviour models of
window control in summer, Build. Environ. 43 (2008) 1471–1482.
[17] F. Haldi, D. Robinson, On the behaviour and adaptation of office occupants,
Build. Environ. 43 (2008) 2163–2177.
[18] H.B. Rijal, P. Tuohy, M.A. Humphreys, J.F. Nicol, A. Samuel, I.A. Raja, J. Clarke,
Development of adaptive algorithms for the operation of windows, fans, and
doors to predict thermal comfort and energy use in Pakistani Buildings,
ASHRAE Trans. 114 (2) (2008) 555–573.
[19] D. Levie, Y. Kluizenaar de, E.C.M. Hoes-van Oeffelen, H. Hofstetter, S.A.
Janssen, M.E. Spiekman, F.G.H. Koene, Determinants of ventilation behaviour
in naturally ventilated dwellings: identification and quantification of
relationships, Build. Environ. 82 (2014) 388–399.
[20] M. Schweiker, F. Haldi, M. Shukuya, D. Robinson, Verification of stochastic
models of window opening behaviour for residential buildings, J. Build.
Perform. Simul. 5 (2012) 55–74.
[21] R. Andersen, V. Fabi, J. Toftum, S.P. Corgnati, B.W. Olesen, Window opening
behaviour modelled from measurements in danish dwellings, Build. Environ.
69 (2013) 101–113.
[22] S. Shi, B. Zhao, Occupants’ interactions with windows in 8 residential
apartments in Beijing and Nanjing, China, Build. Simul. 9 (2016) 221–231.
[23] J.F. Nicol, M.A. Humphreys, Adaptive thermal comfort and sustainable
thermal standards for buildings, Energy Build. 34 (2002) 563–572.
[24] D. Robinson, Some trends and research needs in energy and comfort
prediction, in: Proceedings of Comfort and Energy Use in Buildings-Getting
Them Right Conference, Windsor UK, 2006.