Upper limits for thermal comfort in a passively cooled office environment
across two cooling seasons
Kit Elsworth1, Rod Bates1, Ryan Welch1, and Billie Faircloth1
1 KieranTimberlake, Philadelphia, USA
Abstract: During two summers in a hot humid climate, an architecture firm conducted a thermal comfort study
in a passively cooled office to better the understand the limits of thermal comfort. The office, located in a
renovated industrial building in Philadelphia, relied upon natural ventilation, elevated air movement, and
desiccant dehumidification for cooling. Thermal comfort surveys were sent to the staff and matched to
corresponding ambient temperature and humidity measurements, totalling almost 10,000 survey responses
across a 11.5 °C range of indoor temperatures. The overall findings suggest that 80% of the population was
satisfied at 28.5 °C and 90% at 27.5 °C. Regression models predict thermal comfort based on indoor temperature
and indicate that humidity and clothing did not significantly impact comfort. Occupant clothing insulation (clo)
value decreased with temperature from 21 -27 °C, resulting in a minimum clo value of 0.50, including chair
insulation, for temperatures above 27 °C. The results show strong agreement with the rate of adaptation in
the adaptive thermal comfort model and supports the increase in the comfort threshold given 1.0 m/s of
elevated air movement. The findings from the continuous observations of an occupant population across two
summers allow for validation of the adaptive thermal comfort model and support design strategies that can
maintain comfort over a large range of indoor thermal conditions.
Keywords: thermal comfort, adaptive thermal comfort, field study, longitudinal study, passive cooling
Professional office environments historically operate within a narrow, uniformly applied
temperature band (ASHRAE, 1992) (ISO, 1994). Offices are mechanically conditioned to
ensure temperatures consistently fall within this band throughout all workspaces, including
desks, conference rooms, and support spaces, permitting human bodies to achieve
homeostasis given workplace-appropriate clothing and typical metabolic rates for sedentary
activity. These assumptions are reflected in the first indoor thermal comfort standard
developed in the late-1960s (Fanger, 1967) (Fanger, 1973), which rationalized human thermal
comfort as a heat balance equation based on findings from a comfort chamber study.
Subsequent field studies challenged this equation, finding indoor thermal comfort varies by
climate, culture, and behavior (Humphreys, 1976). This opened a new era of thermal comfort
where researchers sought to show that humans’ ability to thermally adapt is influenced by
various social and behavioral factors, ultimately leading to the development of the ASHRAE
adaptive thermal comfort model (de Dear & Brager, 1998). The adaptive thermal comfort
model states these social and behavioral factors, along with past thermal history, modify the
thermal preferences of building occupants such that individuals’ thermal preferences
depends on season and climate. Research has found building conditions that align with the
adaptive comfort model result in more satisfied occupants under a wider range of indoor
conditions (de Dear & Brager, 1998). Therefore, this standard has become the basis for the
design of passive cooling strategies.
Given the current demand for carbon neutral buildings, architects and engineers have
motivation to evaluate the effectiveness of passive conditioning for professional office
environments (AIA, 2018). Thermal comfort field studies are a crucial method of evaluation
because they capture individual perception and adaptation to real world conditions,
accounting for cultural constraints and diurnal swings - factors difficult to integrate into
controlled climate chamber studies or broad post occupancy evaluations, demonstrating that
the adaptive comfort model better defined comfort. (Nicol & Roaf, 2005). Researchers
previously used field studies to compare the effectiveness of adaptive comfort theory and
passive cooling to the design standards of various climate zones (Nicol, et al., 1999)
(Indraganti & Rao, 2010). Many field studies found that when occupants are allowed to adapt,
they maintain thermal comfort above code-required set points (Karyono, 1995) (Malama, et
al., 1998) (Chan, et al., 1998) (Wagner, et al., 2007) (Luo, et al., 2015).
The multivariate nature of thermal comfort field studies in office environments make
them difficult to execute and interpret. Researchers have limited access to building sites and
survey participants over a prolonged duration. Despite office environments generally
supporting seated workers performing light tasks, the transferability of results can be
challenged by differences in climate zone, building volume or floor area, furniture type and
arrangement, and proximity to building features, such as air diffusers or windows.
Furthermore, employees are often required to adhere to office policies that assist or deter
individual adaptation, such as dress codes, seat assignments, work schedules, and degree of
personal control over personal environment such as the opening or closing of a window or
the use of a desk fan. The totality of these factors must be considered when designing a field
study, collecting data for it, and transforming insights into actions that can be taken by
building designers, owners, and managers.
Here we present results from a longitudinal field study of a passively-conditioned
professional architecture office located in Philadelphia, Pennsylvania (USA). The two-story,
6,300 m2, masonry building was originally constructed in 1949 as a beer bottling factory. A
deep retrofit in 2015 converted the building into an open office with physical and digital
fabrication studios. The main office component is located on the second floor, which features
a 12 m tall ceiling and a clerestory row of north, south, and west facing windows as well as a
north to south central roof monitor. The retrofit intended to obviate the need for mechanical,
chiller-based cooling, which led to the implementation of a passive, naturally ventilated
The building and its inhabitants provide a unique setting, free of typical field study
constraints, for research on passively conditioned professional office environments and
human adaptation. The architecture firm approached the retrofit as an experiment, including
the selected conditioning systems and spatial planning details. The firm anticipated and
planned a multi-season field study which commenced in spring 2015, shortly after occupancy.
A small team of the firm’s architects and researchers assumed responsibility for building
controls and implemented an active management approach. Throughout the study period,
the team collected and interpreted data to inform management protocols, such as advisories
to the building’s inhabitants, logic statement adjustments, and modifications to building
systems. The firm’s office policies encouraged rapid human adaptation during the cooling
season, permitting casual dress, mobility across the office, and flexibility of work schedule.
Participants in the study were incentivized through self-interest and commitment to the
mission of sustainability with the knowledge that the study may advance practices within the
This field study involved occupant surveys and indoor environmental measurements
collected over the summers of 2015 and 2016. The findings are used to address the following
questions: For a professional office environment implementing passive conditioning
strategies, what is a reasonable temperature limit of human thermal comfort? What
percentage of office workers are satisfied, or alternatively dissatisfied, under these
conditions? And which variables contribute to satisfaction and dissatisfaction? This long-term
study offers valuable insight into the experiences of individuals attempting to maintain
comfort during a daily routine. These findings can be shared with practitioners to encourage
experimentation and used to improve the design and implementation of passive strategies in
new construction. These findings may also be used to refine design standards or thermal
comfort models that predict thermal comfort under hot and humid conditions in office
2.1. Building Operations
The building integrates a variety of passive strategies to optimize occupant comfort while
minimizing energy use (Figure 1). The passive strategies include night flushing, stack
ventilation, and natural ventilation. The night flushing process relies on the monitor fans to
draw 17 m3/s of air into the building through the building’s windows on the north, south, and
west facades. Stack ventilation, achieved through closing the office level windows and
exhausting air through the central monitor windows, leverages natural buoyancy and is used
when exterior temperatures, humidity, or dust levels make natural ventilation undesirable.
When exterior conditions are appropriate, natural ventilation is achieved by opening the
windows on the north, south, and west sides of the building envelope. During the cooling
season, locally driven cooling efforts included occupant controlled 1 m diameter industrial
floor fans and 5 Watt personal desk fans capable of moving air at approximately 1.0 m/s at
the occupants’ position. This air speed was the maximum air speed measured with a hot-wire
anemometer (Kanomax A044 0.10-30 ±0.015 m/s) at heights of 0.1 m and 1.1 m with all fans
active. In addition to the passive measures used to cool the building, the office minimizes
solar heat gain throughout the summer via interior roller shades with 10% openness on the
south and west windows.
Figure 1. An east-west building section showing airflow under natural ventilation. The building’s tall open
office allows for stratified air to accumulate at the top of the space and easily exhausted via operable monitor
Mechanical ventilation was delivered via an underfloor air distribution (UFAD) system
with local control by occupants, enabled with a manually actuated swirl diffuser located near
each desk. The UFAD system’s supply air, a mix of outdoor air and indoor return air, was
latently cooled with a 15-ton liquid desiccant dehumidification system. During the study’s first
month, the building underwent a commissioning exercise to determine the best passive
operational sequence, resulting in highly variable interior conditions. The daily sequence of
night flushing followed by natural or stack ventilation, as appropriate, ultimately became the
building’s normal routine during the study period.
The 26-week study period began with 98 participants and by its conclusion grew to 130, 60%
of whom were male and 40% were female. Due to the variations in office attendance,
attributable to hiring, travel, vacation, or illness, the number of participants varied daily.
While the study did not monitor daily occupancy, the accuracy of the study’s calculated
participation rates was improved using the office vacation and travel calendars to account for
the members of the study population known to be out of the office. This method resulted in
a more dynamic and accurate assessment of the study participation than would have been
achieved by assuming the total number of current employees present in the office was
The thermal comfort study was conducted using surveys emailed as a hyperlink to the entire
office staff every weekday at 10:00 AM and 4:00 PM. These times were chosen to fall well
within typical arrival and departure times to capture the greatest number of responses and
avoid any effect of increased metabolic rate resulting from participants’ morning commute.
The web-based survey form (Figure 2) consisted of four questions: Employee ID, Attire,
Thermal Sensation, and Location. Employee IDs were collected to determine each individual’s
weekly response rate and participation across the study period. The Attire question consisted
of a dropdown list of ten representative clothing assemblies from which the participant could
select the outfit that most closely resembled their attire. This list covered of a broad range of
clothing insulation values ranging from a sleeveless dress (0.31) to a jacket, pants, and a long-
sleeve shirt (0.99). The total clo value for each clothing assembly was calculated using the
University of California, Berkeley’s Center for the Built Environment’s Comfort Tool, including
a 0.10 increase in overall clo value to account for insulative value of office chairs (Hoyt, et al.,
2013). In addition to the Attire question’s dropdown list, the Thermal Comfort question was
presented as a radio button list reflecting the Bedford 7-point scale (Bedford, 1936). Lastly,
the Location question took the form of an interactive office floorplan that allowed
participants to select the workstation or room from which they were taking the survey.
Figure 2: Thermal comfort survey form and screenshot of location selection on office floorplan.
Following the initial 26-week study, a separate 10-week survey was administered to
understand the participants’ typical metabolic rates. This second web-based survey was
formatted and administered in a manner similar to the initial survey, but also required
participants to indicate the highest activity level they experienced in the past 20 minutes. This
activity question presented participants with a dropdown list of options representing a variety
of activities including sitting, walking, standing, biking, running, and working in the office’s
2.4. Sensor Hardware
Concurrent with the thermal comfort surveys, a network of sensors recorded the second floor
workspace interior temperature and relative humidity in five-minute intervals at three
locations throughout the main office level (Figure 3). Each sensor consists of an analog
humidity sensor with a ±3% manufacturer’s stated accuracy and a 1-Wire integrated circuit
with an in-chip temperature sensor that has a ±0.5 °C manufacturer’s stated accuracy. Each
sensor is housed in custom 3D-printed casing with a stainless-steel mesh enclosure that
shields the technology from radiation and dust while still permitting airflow.
Figure 3: Floor plan of the office showing location for the three ambient temperature and humidity sensors,
mounted at 1.1 meters above the finished floor.
The sensor locations were selected as representative indicators of the air temperature
and relative humidity in the building’s occupied zones. Since no significant differences
between the readings of these three sensors were found during the study period, recorded
temperatures at each measurement interval were averaged and used to match survey
responses. Consequently, all further use of the term “interior air temperature” refers to the
average of these three simultaneous sensor readings.
During periods in which the sensor network was unable to transmit or record sensor
data, average indoor environmental data from the building management system (BMS) was
used to impute these instances. The BMS communicates with five thermostats located across
the office, all of which meet ASHRAE 55-2013 Section 220.127.116.11 requirements since they record
temperature with an accuracy of ±0.5 °C and relative humidity with an accuracy of ±2%.
Outdoor temperature was also recorded in five-minute intervals by a rooftop-mounted
climate monitoring station (± 1.1 °C at 0.1 °C resolution and ± 5% RH at 1% resolution).
2.5. Statistical Analysis
The Chi Squared test was used to determine significance across the range of environmental
conditions and survey votes. Multiple linear regression was used to fit the data of this study
and produce a formula for comfort given the variables of air temperature, humidity, and
clothing. A logistic regression determined the probability for officewide comfort given the
indoor air temperature. Bootstrap resampling provided cross-validation of the multiple linear
regressions, performing a 100-fold cross validation test for the regression model using a 70%
training set (Kuhn, 2008). For every cross-validation, the Mean Square Error (MSE) and
coefficient of determination (R2) were calculated and the standard deviations were recorded
to understand variances in model performance.
This thermal comfort field study amassed an extensive dataset that includes both survey and
sensor measurements. With a dataset of 9,889 survey responses collected along with indoor
environmental measurements collected between 2015 and 2016, many possible topics of
study emerge. This specific analysis attempts to determine the limits of thermal comfort in a
passively conditioned environment, the factors that influence thermal comfort, and a
reasonable temperature limit for a passively conditioned professional work environment.
3.1. Survey Statistics
To determine whether the survey dataset is representative of the office population, response
and participation rates were calculated on a weekly basis for both summers, with individuals
who responded at least once during the week being counted as participants. Across the whole
study, the average response rate was 39%, of which 47% of the responses were submitted by
females and 53% were submitted by males. Compared to the ratio of females to males in the
office, this result indicates that women chose to respond to the survey more frequently.
Table 1: Survey statistics for 2015 and 2016. Average response rate across the study was 39%, while an
average of 73% of individuals participated on a weekly basis.
Results from the metabolic rate survey showed that 93% of participants were either
sitting (81%) or standing (12%) before or during completion of the survey. Only 7% reported
walking, while less than 1% reported running, biking, or working in the fabrication studio. Due
to the minimal difference in metabolic rates between seated (1.2 MET) and standing work
(1.4 MET) (ASHRAE, 2013) , as well as the large percentage of seated and standing participants
in the survey database (93%), these data were not parsed for metabolic rate.
3.2. Indoor and Outdoor Environmental Conditions
During the study, maximum and minimum daytime indoor air temperatures during occupancy
spanned from 21.5 to 33 °C. Overall, an average day’s temperature gain in the office was
4.5 °C as the building warmed from an average low of 24.5 °C to an average high of 29 °C.
However, the occasional heat wave resulted in several instances where indoor temperature
peaked between 29-31 °C and on occasion eclipsed 32 °C. Corresponding relative humidity
values were between 25% and 80%, translating to a dewpoint range from 4.5-23.5 °C.
Despite fluctuating significantly, the building’s indoor climate did not experience the
same extremes as the outdoor environment. The full range of exterior temperature and
humidity levels depicted in Figure 4 show outdoor temperatures as high as 36 °C and recorded
dewpoints as high as 27 °C. Contributing to this difference in indoor and outdoor
temperatures and humidity levels was the building’s dehumidification system, as well as the
ability to close the building’s windows when outside conditions were warmer than those
Figure 4: Outdoor air temperature and dewpoint distribution for 2015 and 2016.
The average outdoor temperatures and dewpoints between the two years only differed
by 0.05 °C and 0.3 °C, respectively. The small difference in the outdoor temperature and
humidity data between the study’s two summers indicates that both years experienced
similar climatic conditions and produced consistent exterior climatic conditions, allowing the
two summers to be combined for an aggregated analysis on thermal comfort. To ensure the
accuracy of on-site sensor readings, the data recorded by the rooftop-mount climate station
during the study’s two summers was compared, validated, and found to be consistent with
simultaneous data from Philadelphia International Airport.
3.3. Thermal Comfort Profile – Temperature and Humidity
Categorical thermal comfort responses were plotted according to temperature and relative
humidity (Figure 5). The pronounced lateral gradient in this scatter plot suggests temperature
was the primary factor influencing participants’ thermal comfort, while the absence of a
vertical gradient suggests that relative humidity had little effect on comfort. For example,
survey responses corresponding to indoor temperatures between 25-29 °C show participants
were comfortable regardless of relative humidity levels, which ranged from 25-75%.
Humidity’s limited influence on participants’ perceived comfort is most likely due to the
office’s increased elevated air movement, an explanation that is consistent with previous
research that shows air movement to increase occupants’ humidity tolerance (Zhai, et al.,
2015) (Melikov, et al., 2008). Furthermore, the European standard EN15251 found that
humidity has a small impact on thermal sensation when developing the European adaptive
thermal comfort model (Nicol & Humphreys, 2010).
Figure 5: All survey responses against corresponding indoor temperature and relative humidity.
To determine the relationship of temperature, humidity, and clothing on thermal
comfort, a multiple linear regression was produced. Each variable was centered and scaled to
allow for weighting and comparison of all predictor variables. These results are shown in Table
Table 2: Comparison of coefficients from centered and scaled multiple linear regression.
Indoor Temperature (°C)
p < 0.001
Indoor Dewpoint (°C)
p < 0.001
p = 0.001
p < 0.001
The regression equation for thermal comfort using temperature, dewpoint, and clo is
summarized in Equation 1, where TC represents the seven-point thermal comfort vote, Tin
represents the indoor air temperature (°C), Tdp represents the indoor dewpoint (°C), and clo
represents the respondent’s clothing insulation value. For all predictor variables, the standard
error was less than 0.009 and p-value less than 0.001. The regression coefficients in Table 2
validate the belief that temperature has the greatest impact on comfort due the large
weighting of the coefficient with respect to the other predictor variables. The small coefficient
corresponding to indoor dewpoint reinforces the belief demonstrated in Figure 5 that
dewpoint does not greatly affect thermal comfort. It is surprising that an increase in clo value
will decrease the thermal comfort vote, suggesting that occupants were more comfortable
with more layers of clothing. This result may be influenced by the discomfort experienced by
individuals wearing low levels of clothing on hot days, thus experiencing discomfort due to
indoor conditions and not clothing choice. The negative clothing coefficient may reveal that
a general correlation exists with occupant comfort and higher clothing levels at hotter
temperatures. However, this should not be considered causation since the small coefficient
suggests the impact is minimal, estimating an increase of 0.1 clo results in a 0.04 decrease in
thermal comfort vote.
The regression model in Equation 1 resulted in MSE of 0.58 and a R2 of 0.46, which
match with the statistics of a cross-validated model, showing a strong representation of
typical conditions. The 100-fold cross-validation yielded an average MSE of 0.58 ±0.008 and
R2 of 0.47 ±0.010. The low standard deviation reflected in this figure indicates the model has
a low variance and high repeatability.
3.4. Thermal Comfort Profile – Population Satisfaction
Comfort is defined as thermal comfort votes including Cool but Comfortable, Neutral, and
Warm but Comfortable. More than 95% of the office population expressed comfort at
temperatures below 27 °C (Figure 6). Between 27 and 32 °C, the population’s comfort
decreased somewhat linearly, demonstrating temperature’s significant effect on thermal
comfort (chi squared p<0.05). When temperatures exceeded 31.5 °C, thermal comfort did not
vary beyond 20-30% of the office being comfortable.
ASHRAE 55-2013 recommends a building’s mechanical systems be designed to provide
80% of the population with comfortable conditions. Based on the results of this study, a
building owner operating a building with similar passive measures wishing to adhere to this
threshold should aim for a setpoint temperature of 28.5 °C. Reducing this average
temperature by only 1 °C to 27.5 °C, however, would result in 90% occupant comfort.
Additionally, air temperatures below 22 °C may cause cool discomfort.
Figure 6: Less less than 80% of the population was comfortable when temperatures exceed 28.5 °C.
The comfort votes were converted into a binary variable to perform a logistic regression
to determine the statistical structure of the comfort curve due to warm discomfort.
Where Tin as indoor air temperature (°C) and Pc as the probability that the vote is
comfortable. The logistic regression represented in Equation 2 can be used to accurately
predict the probability for comfort within the population given the indoor air temperature in
the passive environment studied here. The curve of this regression roughly follows the path
of the black lines in Figure 6.
Equation 1 shows clo value has minimal effect on thermal comfort but examining clo value
against temperature offer insights into the choices of clothing across indoor conditions. Over
the course of the study, the average clo value was 0.51 ±0.07 (N=9,889), including the 0.10
clo added for chair insulation to all responses. These clo values are on the lower end of other
longitudinal case studies, which found average summer clo to range between 0.5 and 0.55
(Schiavon & Lee, 2013) (Honnekeri & Pigman, 2014). Removing the contribution of the chair,
the actual clo value for the clothing assembly alone produces a median value of 0.41. This
value is more indicative of the typical clo value of occupants observing a flexible dress code
in an office environment in the summer.
The participants’ clo values were plotted against indoor temperature, showing only
votes indicating comfort to remove any instances when clothing, rather than temperature or
humidity, may have been the source of discomfort (Figure 7). This same graph also includes
the average clo value during each recorded indoor air temperature, as well as a 95%
confidence curve. The graph’s regression line shows that the rate of clo value decline
decreases from 21 °C to 27 °C. Above 27 °C, a clo value of 0.50 was maintained, exhibiting the
minimum clo value preferred in this office setting. Interestingly, this regression curve
performs opposite in nature to the decrease in population comfort in Figure 6. While clo
values decline until 27 °C, the percentage of comfortable participants in Figure 6 remains
constant. Above 27 °C, clo values reach a steady minimum value as comfort percentage
decreases. It suggests that 27 °C serves as an inflection point for both clothing and overall
satisfaction, a trend that indicates that, above 27 °C, individuals are not able to reduce
clothing and dissatisfaction will increase.
Figure 7: The average clo values decrease until temperatures reach 27 °C. At temperatures above 27 °C, clo
values reach a steady minimum value of 0.50.
4.1. Comparison with the Adaptive Comfort Model
Following the study’s analysis, the authors calculated the optimal indoor air temperature per
prevailing outdoor mean temperature, defined as the arithmetic average of the previous
seven days’ mean daily outdoor temperature. The purpose of this calculation was to compare
survey results to the ASHRAE adaptive thermal comfort model. Despite not having the means
to measure globe temperature during the study, spot measurements across the study’s two
seasons with a Kestrel 5400 Heat Stress Meter (accuracy 0.5°C dry-bulb, 1.5°C globe
temperature) found minimal difference between the globe temperature and indoor air
temperature. Consequently, indoor air temperature rather than operative temperature was
used during the adaptive comfort calculation.
While it is simple and straightforward to calculate the neutral temperature, or optimal
comfort temperature, of this study (26.6 °C), this single metric does not capture the full range
of temperatures at which a population will feel thermally neutral. As described by Nicol and
Humphreys (Nicol & Humphreys, 2010), the neutral temperature in a variable climate
becomes a “moving target” and changes as individuals adapt to their environment. Therefore,
regressions to determine neutral temperature based on prevailing outdoor mean
temperature dataset are provided in Equation 3 (R2=0.50).
KieranTimberlake Study: (3)
ASHRAE Adaptive Comfort Model:
(Schilling Brager & de Dear, 2000)
Where Top is operative temperature (˚C) and Tout-mean is the seven-day average outdoor
monthly mean temperature (˚C). When compared with the regression used to define the
adaptive thermal comfort zone, the coefficient for operative temperature differs only by 0.05.
The equivalency of the outdoor monthly mean slopes suggests the adaptive thermal
comfort zone predicts that humans will adapt to warmer temperatures to a greater extent
than the individuals in this study. This may be due to the influence of a global population on
the ASHRAE adaptive thermal comfort model rather than a sample from a single city.
However, due to the intercept, this study’s regression line trends slightly above the adaptive
comfort zone’s neutral temperature, indicating a greater threshold of comfort (Figure 8). The
offset is due to the intercept value from this dataset, which occurs approximately 2.2 ˚C above
the adaptive comfort zone’s neutral temperature when outdoor mean is equal to zero. Within
the temperatures experienced in the indoor environment, this offset is between 1.0-1.8 °C .
This increase in the upper limits of comfort is similar to the increased threshold for comfort
allowed by the adaptive comfort model when 0.9 m/s of air movement is provided, as per
ASHRAE Standard 55-2013. This air speed is found to be roughly equal to the 1.0 m/s of air
movement experienced at the occupant’s seated position under the influence of fans and can
be seen as evidence to validate the increase in thermal comfort preference attributed to
elevated air speeds of approximately 1.0 m/s.
Figure 8: A graph plotting the optimal indoor temperature for a range of outdoor prevailing mean
temperatures showings the study’s comfort limits are approximately 1.0-1.8 °C greater than the adaptive
thermal comfort model.
Although this study only allowed for the submission of discrete clo values, the results
showed that the rate of clo value declined slower than predicted in the adaptive clothing
model. The adaptive thermal comfort model estimates the median clo value will decrease 0.1
clo for every 2 °C increase in indoor temperature (de Dear & Brager, 1998). In this study, the
rate of decline in clo value between 21 and 27 °C was found to decrease at a rate of 0.1 clo
per 4.5°C indoor air temperature. This rate declined as temperature increased as occupants
gradually reached the minimum clo value, 0.5, experienced in the office.
4.2. Measurement Limitations
It should be recognized that this study is not without limitations, but these limitations are
opportunities to improve future data collection. For one, it was not possible to continuously
monitor air speed during the study. Instead, air speeds were measured using an anemometer
spot measurements. While these measurements were representative of the predominant air
speed felt by survey participants, it is possible that individuals seated near windows or close
to floor fans experienced a greater air velocity.
Globe temperature was also not measured continuously. Although the authors are
confident that the MRT based on envelope and equipment gains was approximately equal to
the building’s indoor temperature, the impact of direct solar was not captured. The building’s
direct sun exposure through monitor windows is transient on any given day, but its presence
may have overlapped with the time in which participants completed their daily surveys.
Lastly, one of the greatest challenges during data collection was tracking the survey’s
participants. Without the ability to track daily occupancy, the office’s population was
estimated each week using office-wide vacation and travel calendars. A more accurate
method of tracking building occupancy will improve the calculation of survey participation
rate for future research.
4.3. Examining Standard Survey Metrics
During this study, the authors discovered constant communication with survey participants
was the most effective way to maintain survey participation. However, the study also found
that participants’ interest declined over the study period, indicating some level of survey
fatigue. ASHRAE 55 suggests using response rate to determine the efficacy of a survey and
whether the survey reached a representative sample size. The challenge with relying on
response rate as a measure of success is that response rate is not indicative of nonresponse
bias or survey fatigue, both of which are additional metrics that longitudinal surveys should
consider. Likewise, closely tracking survey fatigue allows future researchers to decide
whether to remind participants to complete surveys or to conclude their study. In addition,
administrators conducting longitudinal surveys should understand nonresponse bias not only
in individuals but in spatial distribution. A spatially distributed set of responses across a
building will ensure that the dataset is not representative of a single zone or floor.
ASHRAE 55-2013 provides recommended response rates for thermal comfort studies
based on three population sizes. For populations greater than 45 people, response rates
should exceed 35%. However, this standard does not differentiate between transverse and
longitudinal surveys, nor does it provide guidance on calculation methodology. Other
standards, such as ASHRAE’s Performance Measuring Protocol (ASHRAE, 2010) dictate a 40%
response rate. In both instances, the calculation methodology is not disclosed in the
standards, an omission that invites a generous interpretation of this survey metric. This room
for interpretation is significant, especially when considering that calculating response rates
on a daily, weekly, monthly, or cumulative basis can yield significantly different results. When
conducting a study-wide analysis, longitudinal surveys with weeks of low participation can
seem valid by accounting for a high response rate at the outset of the study. Alternatively,
tracking individual response rates at a granular scale, for example per survey issuance or by
day, can allow the survey dataset to be reliably normalized and may justify exclusion of survey
responses that are not representative of the whole population.
There is an opportunity to improve longitudinal survey protocol and analysis with data
driven findings. The survey fatigue experienced in the 2016 portion of this study offers a
unique opportunity to identify the minimum required response rate or test other analytical
methods that can produce similar, statistically significant results. In addition, resampling the
dataset may also serve as another means to artificially lower the response rate and examine
its effect on the study’s outcome. In this study, the close agreement of MSE and R2 between
the regression and the resampling along with the low standard deviation indicated a high
repeatability and demonstrated confidence that the survey dataset did not suffer from
Over the course of two summers in a hot and humid climate, 90% of the population was
comfortable at 27.5 °C in a passively cooled office environment. When indoor air
temperatures rose to 28.5 °C, 80% of the building’s population remained thermally
comfortable. Indoor air temperature was found to be the most influential predictor of
comfort, while humidity and clothing had little impact. Participants’ clo values decreased in
response to temperature until indoor temperatures reached 27 °C, at which point the average
clo value maintained 0.5 clo.
This field study validates the adaptive thermal comfort model under elevated air
movement by demonstrating that the thermal neutrality line occurs 1.0-1.8 °C greater than
the adaptive comfort model when providing 1.0 m/s of air speed to occupants. These results
confirm the applicability of the adaptive thermal comfort model in passively cooled modern
office buildings where occupants have operable windows and access to fans.
Acknowledging that research on such high indoor temperatures and humidity levels is
difficult to conduct, it is the authors’ intent for this study to demonstrate the full range of
thermal comfort in a passively cooled office space in a hot and humid climate. Although the
design of the study could not quite achieve 100% comfort during the summer, the ability to
provide passive cooling for a majority of the season may encourage designers to move toward
mixed-mode designs that reduce reliance on air conditioning within reasonable limits. In
addition, the authors hope this study will empower building owners to survey their own
populations, using the resulting data to determine how best to reduce cooling demand while
maintaining occupant comfort.
This study was the vision of Stephen Kieran and James Timberlake, who provided the
facility and access to staff, in addition to guidance and encouragement. The KieranTimberlake
staff deserve recognition for their participation in the surveying to research human thermal
comfort in a contemporary office setting. Jason Niebish and Paul Worrell deserve credit for
their hard work managing the building during this study. Alex Knipe was instrumental to the
survey’s development. The authors would also like to acknowledge the support of Ed Arens
and Stefano Schiavon from the University of California, Berkeley’s Center for the Built
Environment for their insightful comments and suggestions.
AIA, 2018. The 2030 Committment. [Online]
Available at: https://www.aia.org/resources/6616-the-2030-commitment
[Accessed 12 January 2018].
ASHRAE, 1992. ANSI/ASHRAE Standard 55-2013, Thermal Environmental Conditions for Human Occupancy,
Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers.
ASHRAE, 2010. Performance Measurement Protocol for Commercial Buildings. Atlanta: American Society of
Heating, Refrigerating and Air Condition Engineers.
ASHRAE, 2013. ANSI/ASHRAE Standard 55-2013, Thermal Environmental Conditions for Human Occupancy,
Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers.
Bedford, T., 1936. The Warmth Factor in Comfort at Work. A Physiological Study of Heating and Ventilation.
London: Medical Research Council Industrial Health Research Board.
Chan, D., Burnett, T., de Dear, R. & Ng, S., 1998. A large scale survey of thermal comfort in office premises in
Hong Kong. ASHRAE Technical Data Bulletin, Volume 14, pp. 76-84.
de Dear, R. & Brager, G., 1998. Developing an adaptive model of thermal comfort and preference. ASHRAE
Transactions, 104(1), pp. 145-167.
Fanger, P., 1967. Calculation of thermal comfort, Introduction of a basic comfort equation. ASHRAE
Transactions, Volume 73, pp. 4.1-4.20.
Fanger, P., 1973. Assessment of man's thermal comfort in practice. Occupational and Environmental Medicine,
30(4), pp. 313-324.
Honnekeri, A. & Pigman, M., 2014. Use of Adaptive Actions and Thermal Comfort in a Naturally Ventilated
Office. IndoorAir Conference PrePrint.
Hoyt, T., Schiavon, S., Piccioli, A., Cheung, T., Moon, D. & Steinfeld, K., 2013. CBE Thermal Comfort Tool.
Available at: http://cbe.berkeley.edu/comforttool/
Humphreys, M., 1976. Field Studies of Thermal Comfort Compared and Applied. Journal of the Institution of
Heating and Ventilating Engineers, Volume 44, pp. 5-27.
Indraganti, M. & Rao, K. D., 2010. Effect of age, gender, economic group and tenure on thermal comfort: A
field study in residential buildings in hot and dry climate with seasonal variations. Energy and Buildings,
42(3), pp. 273-281.
ISO, 1994. International Standard 7730, Moderate thermal environments - Determination of the PMV and PPD
indices and specifications of the conditions of Thermal Comfort, Geneva: International Standards
Karyono, T. H., 1995. Thermal comfort for the Indonesian workers in Jakarta. Building Research and
Information, 23(6), pp. 317-323.
Luo, M. et al., 2015. Evaluating thermal comfort in mixed-mode buildings: A field study in a subtropical
climate. Building and Environment, Volume 88, pp. 46-54.
Malama, A., Sharples, S., Pitts, A. & Jitkhajornwanich, K., 1998. Investigation of the thermal comfort adaptive
model in a tropical upland climate. ASHRAE Transactions, Volume 104, pp. 1194-1206.
Melikov, A., Kaczmarcyzk, J. & Sliva, D., 2008. Impact of air movement on perceived air quality at different level
of relative humidity. IndoorAir. Copenhagen. 2008., pp. 17-22.
Nicol, F. & Humphreys, M., 2010. Derivation of the adaptive equations for thermal comfort in free-running
buildings in European standard EN15251. Building and Environment, Volume 45, pp. 11-17.
Nicol, F., Raja, I., Allaudin, A. & Jamy, G. N., 1999. Climatic variations in comfortable temperatures: the
Pakistan project. Energy and Buildings, 30(3), pp. 261-279.
Nicol, F. & Roaf, S., 2005. Post-occupancy evaluation and field studies of thermal comfort. Building Research
and Information, 33(4), pp. 338-346.
Schiavon, S. & Lee, K. H., 2013. Dynamic predictive clothing insulation models based on outdoor air and indoor
operative temperatures. Building and Environment, Volume 59, pp. 250-260.
Schilling Brager, G. & de Dear, R., 2000. A Standard for Natural Ventilation. ASHRAE Journal.
Wagner, A. et al., 2007. Thermal comfort and workplace occupant satisfaction—Results of field studies in
German low energy office buildings. Energy and Buildings, 39(7), pp. 758-769.
Zhai, Y. et al., 2015. Human comfort and perceived air quality in warm and humid environments with ceiling
fans. Building and Environment, Volume 90, pp. 178-185.