Building and Environment 92 (2015) 396e406

Contents lists available at ScienceDirect

Building and Environment

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

A thermal comfort eld study of naturally ventilated classrooms in

Kharagpur, India
Asit Kumar Mishra*, Maddali Ramgopal
Department of Mechanical Engineering, Indian Institute of Technology, Kharagpur 721302, India

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 3 February 2015
Received in revised form
27 April 2015
Accepted 21 May 2015
Available online 27 May 2015

To assess occupant thermal comfort, eld studies were carried out in naturally ventilated (NV) classrooms of Indian Institute of Technology Kharagpur. The location has a hotehumid climate. Surveys were
taken during both semesters over the academic year 2013e14. Results of the surveys gave a regression
neutral temperature near 29
C while preferred temperature was found to be 26.8
C. Using student
responses to thermal acceptability question, 80% occupant satisfaction was found between 22.1 and
31.5
C operative temperature. Over the survey duration, nearly 79% of responses accepted their thermal
environment. Analysis of thermal preference and thermal acceptability votes showed a distinct preference amongst occupants for cooler than neutral sensation. Diurnal variation of temperature that would
be acceptable to 80% or more occupants was found to be a 4
C wide band. Study of student actions
during surveys showed that fans were brought into play more often than windows. Variation of clothing
showed strongest correlation with the day's minimum temperature. Overall, observations from the study
showed broad comfort zones and signicant level of occupant adaptation to the environment of NV
classrooms.
2015 Elsevier Ltd. All rights reserved.

Keywords:
Adaptive thermal comfort
Classrooms
Natural ventilation
Hotehumid climate
Adaptive opportunities

1. Introduction
With 14.6 million enrolled students, higher education system of
India is one of the largest globally. Over the past years, number of
higher education institutes in India have seen compounded annual
growth rate (CAGR) of 11% while student enrolment CAGR was 6%
[1]. Taking these numbers as an indication, it is easy to see that
coming years will witness a dramatic growth in higher education
sector of India. This will mean signicant growth in the sheer
number of classrooms. Conscientious design of classroom thermal
environment is necessary both because of the high occupant densities classrooms have and the adverse impact decient thermal
settings can have on the teaching-learning process. Judicious
comfort standards will be an essential part of designing thermally
comfortable and energy efcient classrooms. Formulation of such
standards, that supplement Indian building codes and pave the way
towards sustainable future development of India, would be

* Corresponding author.
E-mail addresses: asit.mishra@mech.iitkgp.ernet.in (A.K. Mishra), ramg@mech.
iitkgp.ernet.in (M. Ramgopal).
http://dx.doi.org/10.1016/j.buildenv.2015.05.024
0360-1323/ 2015 Elsevier Ltd. All rights reserved.

immensely aided by the results of thermal comfort eld studies.

Over the past couple of decades, a few studies have been carried
out in naturally ventilated (NV) classrooms located in tropics [2e6].
Results of these studies showed that students adapted well to their
NV classrooms. Comfort zones for these surveys were between 24
and 31
C. Looking at the Indian context, Pellegrino et al. [7] carried
out a short duration comfort study for NV classrooms in a couple of
universities located in Kolkata. In two of our earlier works, we have
reported ndings from thermal comfort study for undergraduate
laboratories [8,9]. To the best of our knowledge, no long term
thermal comfort eld study has been reported from Indian classrooms. Thus, to address this gap, the current study was undertaken
with the following aims:
 To obtain year round data regarding occupant thermal sensation
and acceptability for naturally ventilated classrooms in India
 To study student adaptive behaviour during classes and make
observations on how such behaviour helps or hinders thermal
adaptation
 To compare results from the classroom survey with those from
the laboratory [8]; subjects in the laboratory had an appreciably
higher metabolic rate

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

Nomenclature

Abbreviations
AC
air-conditioning
APD
actual percentage dissatised
DBT
dry bulb temperature
MRT
mean radiant temperature
MTSV
mean thermal sensation vote
NV
naturally ventilated
PMOAT prevailing mean outdoor air temperature
PS
percentage of acceptable votes/percentage satised
RH
relative humidity
RMT
running mean temperature
TSV
thermal sensation vote (individual's)
Symbols
pw
ta
tc
tg
tmrt
top
trmt
va
DTc

partial pressure of water vapour in air, kPa

air temperature,
C
comfort temperature, using Grifths' equation,
C
globe temperature,
C
mean radiant temperature,
C
operative temperature,
C
running mean temperature,
C
air velocity, m/s
(top  tc ),
C

2. Methodology
Study was conducted in undergraduate classrooms of Indian
Institute of Technology Kharagpur (IIT). Kharagpur, a small township in eastern India, has a tropical climate with dry winters e
ppen climate classication Aw. The region primarily experiKo
ences three seasons, a hot summer with some sporadic rain, a warm
and humid monsoon and a mild and dry winter. Due to the inuence of south-west monsoon, the bulk of rainfall occurs over the
months of July, August, September, and part of June and October.
Surveys were of longitudinal design and followed one course
each during Autumn 2013 and Spring 2014 respectively. Both
courses were taught in the same classroom. Spring semester in IIT is
from January to April while the Autumn semester is from mid-July
to November. The Autumn semester, unlike the Spring has a midsemester break of about 10 days. The survey took place on 5 days
during Autumn of 2013 and 7 days during Spring semester of 2014.
2.1. The classroom and the subjects
The room (room code: CRe310) is on the top oor of a building
whose major axis is along EasteWest direction. All the rooms in the
top oor of this building are used as classrooms. The EasteWest
orientation of the major axis of the building and the complete top
oor being utilized as classrooms is a feature shared by several of
the neighbouring departmental buildings as well. The building has
pillared construction with brick and mortar lling. Doors of CRe310
open into a corridor while the walls on east and west are internal
walls. Windows are on north faade of the room and are single
glazed with mild steel frames. Windows are 1.8 m wide and 1.5 m
high, with continuous overhangs about one meter deep. Dimensions of CRe310 are as follows: ceiling height 3.4 m, length
6.7 m, breadth 8.3 m Fig. 1 gives the room layout and an idea on
room furnishings.
The survey followed two courses taken in CRe310 e one during

397

Autumn and one during Spring. The Autumn course had 31 students enrolled while the Spring course had 51 students. Fifteen
students were common to both courses, putting the number of
unique subjects at 67. The student population was a mix of undergraduate and graduate students, with age between 19 and 26
years. All subjects were Indians and were assumed to be acclimatized to the local climate. Every effort was made to minimally
inconvenience the classes during which surveys were conducted.
To this end, at the beginning of each semester, students were briey
introduced to the nature and purpose of the surveys. What was
particularly impressed upon the students was that they should feel
free to provide their forthright responses and that the survey was
not asking for a mandate regarding usage of air-conditioning in the
classrooms. Towards later survey days during each semester, in a
behaviour eerily similar to that observed by Teli et al. [11] in their
study with primary school children, some students queried as to
why they had to ll in the same questionnaire repeatedly.
2.2. Collection of survey data
2.2.1. Indoor environmental parameters
Information regarding instruments used for measuring indoor
environmental parameters is given in Table 1. The black globe
thermometer was made in-house by placing an alcohol thermometer at the centre of a 70 mm diameter plastic ball. The ball
was painted black and coated with lamp black. This globe thermometer had also been used in our previous studies [8,12].
For the classes during which surveys were conducted, measurements were taken during the last 20 min of the class. Paper
questionnaires were given to students for lling up during the last
ten minutes. Since the classes were either one or two hours long,
the subjects thus had ample time to attain a stable metabolic rate.
Measurements were taken at six points around the room for air
temperature (ta), relative humidity (RH), and air velocity (va). Globe
temperature (tg) was measured at one central location in the room.
Approximate measurement locations are specied in Fig. 1. All
measurements were taken at approximately shoulder level of occupants. While measuring air velocity inside the classroom, it was
noted that little to no wind came in through windows. This may be
ascribed to the building being surrounded by other buildings and
tree lining. Thus the air velocity measured was almost solely due to
the fans. When the fans were not used, no perceptible air velocity
was recorded. From the ta and RH values, partial pressure of water
vapour in air (pw) was calculated using the Vaisala Humidity
Calculator 3.0 [13]. Values of ta, pw, and va recorded across the six
points were later averaged and used for all further analysis.
Mean radiant temperature (tmrt) was calculated from Equation
(1) [14].

tmrt



tg 273

4



1:1  108 v0:6
a
 tg  ta
D0:4

1=4
 273

(1)

In the above equation, is surface emissivity for the globe and

was taken as 0.95 while D is diameter of the globe in meters.
Room operative temperature (top) was calculated using the
r ,tmrt
[14]. A constant value of 4.7 W/m2,
C
formula: top hc ,tha h
h
c

[14] was assumed for the radiative heat transfer coefcient (hr).
Convective heat transfer coefcient (hc) was calculated from the
correlations for persons seated with moving air in ASHRAE Fundamentals Handbook (Table 6, Chapter 9) [14].
2.2.2. Subjective questionnaire
English, being medium of instruction in all courses, was chosen
to frame the questionnaire. An attempt was made to keep

398

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

Fig. 1. Classroom layout. Points 1e6 are locations where ta, RH, and va were measured. GT gives location of globe thermometer during surveys.

Table 1
Survey instruments.
Instrument and make
Globe thermometer fabricated in-house
Lutron LM 8102,
5-in-1 meter

Measured parameter

Range

Resolution

Accuracy

Air temperature
Relative humidity (RH)
Air velocity
Sound level

10 e 110
C
0e50
C
10e95%
0.4e30 m/s
35e130 dB(A)

1
C
0.1
C
0.1% RH
0.1 m/s
0.1 dB

1
C
1
C
4% RH
3%
3.5 dB

questionnaires brief so as not to take up too much from the class

time. Fig. 2 shows the subjective questionnaire. In addition to
questions regarding how the students perceived their thermal
environment, a short section regarding clothing was also included.

In this section, the most commonly used clothing ensembles

amongst students were listed out and the responder only had to
tick mark the appropriate choice. Numbers given along comfort
questionnaire represent the values used for different responses

Fig. 2. Survey questionnaire.

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

during later analysis. Numbers next to the clothing ensembles

represent the approximate clo values for each ensemble and
include contribution from undergarments and foot wear. These
numbers have been used only for this illustration and were not part
of the original questions.
During winter days, students were asked to write down any
extra winter garment they were wearing next to their clothing
ensemble. For winter garments, clo values were taken from the
work by Nicol et al. [15] in Pakistan: cardigan, 0.22 clo; jacket, 0.30
clo; sweater, 0.30 clo; sweat-shirt, 0.34 clo; waistcoat/vest, 0.18 clo.
2.2.3. Outdoor temperature data
Daily maximum and minimum temperature and precipitation
records were taken from the institute in-campus weather station.
Daily mean temperature (tm) was taken as average of daily
maximum and minimum temperature. Running mean temperature
of day n (trmt) was calculated based on tm values for the last seven
days, using Equation (2) [16]:

n
o
trmt 1  a tm;n1 atm;n2 a2 tm;n3 /

(2)

The value of a was taken to be 0.8, same as the value used for
calculating trmt in EN15251.
Prevailing mean outdoor air temperature, the recommended
outdoor temperature index as per ASHRAE Standard 55 [17], on a
survey day was taken as average of tm from last seven days.
A series of transverse thermal comfort surveys had been carried
out in classrooms near CRe310 for three days during November
2012 and four days during March 2013 [12]. Though the survey
methodology followed was similar, these earlier surveys (Surveys
2012e13) had a slightly different format of subjective questionnaire
(Fig. 3). So, only the thermal sensation votes and thermal acceptability votes of both surveys were analysed together.
During the course of analysis, whenever day-wise study is done
e say for mean thermal sensation vote (MTSV) or acceptability
levels e a weighting scheme is used to take into account the difference in number of responses across survey days. Humphreys
et al. [18] observed from their meta-analysis of thermal comfort
survey data that with about 20 observations, the standard error of
estimated comfort temperature was only 0.4 K. Hence, 20 responses was taken as an upper limit for ensuring desirable accuracy
of results. A weighting factor of one was used for days with
number of responses greater than or equal to 20. Days when
number of responses were less than 20 were assigned a weight of

399

the actual number of responses divided by 20. Including Surveys

2012e13 and 2013e14, there were a total of ve days when less
than 20 responses were collected. All statistical analyses were
carried out with the R statistical computing package [19].
3. Results and discussions
Summary of responses received during Surveys of 2012e13 and
2013e14 are presented in Table 2. Responses were deemed to be
invalid under two circumstances:
 if a subject found her/his thermal environment unacceptable
and yet voted for no change in the thermal preference question
 if a subject voted an extreme thermal sensation (3) and again
voted for more of the same sensation in the thermal preference
question
Only valid responses were analysed further.
Responses from Surveys 2012e13 were taken to be satised if
they voted Acceptable to the question on acceptability of temperature (Fig. 3). On the other hand, a response from Surveys
2013e14 was assumed to be satised if it voted Just Acceptable or
better to the question on acceptability of thermal environment
(Fig. 2).
Maximum and minimum values of some of the important indoor and outdoor conditions during survey are given in Table 3. The
values presented consider observations recorded during both surveys (2012e13 and 2013e14) but due to the longer length of the
2013-14 survey, nal values in Table 3 are all from the later survey's
duration.
3.1. Analysis of thermal sensation votes
A day-wise scatter plot of thermal sensation votes, with some
articial jitter added to distinguish between overlapping votes, is
given in Fig. 4 (a). Due to variability of individual perception, votes
on the thermal sensation scale were quite widely distributed even
on the same day. Average standard deviation of TSV was 0.9, with a
minimum value of 0.6 and maximum of 1.2.
3.1.1. Regression neutral temperature
Regression of TSV with respect to operative temperature was
initially done separately for the 2012e13 and 2013-14 studies.
Resulting relations are given in Equation (3a) and Equation (3b)

Fig. 3. Survey questionnaire for earlier surveys (Surveys 2012e13).

400

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

Table 2
Survey responses.

Survey days
Total responses
Responses from female subjects
Invalid responses
Valid responses
Responses accepting thermal conditions

Table 3
Maximum and minimum values of indoor and outdoor parameters.
Parameter

Maximum

Meteorological data
Daily max
39
C
Daily mean
32.5
C
PMOAT
32.5
C
Indoors data
top
35
C
va
1.2 m/s
MTSV
2

Minimum

Parameter

Maximum

Minimum

19
C
14.5
C
16.6
C

Daily min
RMT

26
C
31.6
C

10
C
16.6
C

20
C
0.05 m/s
1.8

pw
Mean clo
APD

3.55 kPa
0.70 clo
56.8%

1.06 kPa
0.50 clo
0%

2012e13 surveys

2013e14 surveys

7
183
14
7
176
145 (82.4%)

12
365
19
8
357
276 (77.3%)

respectively. Because of the shorter duration and narrower range of

outdoor conditions of the 2012-13 surveys, Equation (3a) has a
higher slope than Equation (3b). Occupant adaptation was more
completely expressed during the 2013-14 surveys. Both regression
relations give a neutral temperature of 29.8
C. This supports our
decision to analyse TSV values from both surveys together.

TSV 0:36top  10:73;

Fig. 4. Thermal sensation votes a) scatter plot b) regression t.

R2 0:35;

p < 0:001

(3a)

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

TSV 0:20top  5:96;

R2 0:43;

p < 0:001

(3b)

Regression relation for TSV with top from both surveys taken
together is given in Equation (4). From this relation, regression
neutral temperature was found to be 29.0
C e which is close to the
results we got from analysing the surveys separately. Also, from
Equation (4), the zone corresponding to votes of 1 was between
24.4 and 33.5
C.

TSV 0:22top  6:37;

R2 0:40;

p < 0:001

(4)

Regression between top and MTSV is given in Equation (5) and

this gives a neutral temperature of 29.5
C. Both Eqns. (4) and (5)
are very similar in form too. Due to inter-individual variation in
how thermal environments were perceived, the regression relation
between top and TSV has a smaller R2 value than the relation between top and MTSV votes. We had a similar nding from our study
in NV laboratories [8]. However, it may be noted here that for
surveys dealing with human behaviour, an R2 value of 0.40 still
indicates a strong correlation [20].

MTSV 0:22top  6:5;

R2 0:73;

p < 0:001

(5)

The neutral temperature found from our surveys is close to the

value obtained in a couple of studies done in naturally ventilated
urban residential buildings, in similar climatic regions d 28.4
C for
Mallick [21] and 29.23
C for Indraganti [22]. Three studies done in
school classrooms in tropical and sub-tropical regions also gave
regression neutral temperatures close to that found in the current
study: 28.8
C for Wong and Khoo [3], 29.2
C for Liang et al. [23],
and 28.4
C Hussein et al. [5]. More interestingly though, a similar
result came unexpectedly from air-conditioned high school classrooms in Taiwan [24]. The air-conditioners in these classrooms
were under occupant control and occupants also had to pay for the
usage. The mean temperature maintained in these classrooms had
been found to be 29.3
C.
3.1.2. Comfort temperature using Grifths equation
In addition to the regression neutral temperature, an average
study comfort temperature (tc) was also calculated using Grifths'
equation: tc top  TSV
m . Table 4 gives the mean values for entire
survey duration of TSV (Survey MTSV) and operative temperature
(top ). From these values, tc is calculated using Grifths' equation. A
slope (m) of 0.5 is used in Grifths' equation in line with the
ndings of Humphreys et al. [18]. As can be seen from Table 4, tc
results for the surveys taken separately as well as combined
together were about equal. This nding again justies the choice of
analysing TSV votes from the two surveys together. At the same
time, the tc value found using Grifths equation is also close to the
regression neutral temperature.

401

 top, r 0.45, p < 0.001

 pw, r 0.31, p < 0.001
 RH, r 0.004, p 0.94
As may be seen, sweating sensation had a reasonable relation
with humidity stated in terms of pw but a non-existent relation
with humidity stated in terms of RH. This nding suggests, as also
found in some earlier works [9,25], that pw is a better indicator of
occupant comfort level with humidity than RH.
3.2. Acceptability of thermal conditions
For free running buildings, as indoor conditions vary with the
outdoors, the range of temperature over which occupants are
comfortable is more important than a single neutral temperature.
Taking data from both 2012e13 and 2013-14 surveys, 79% of the
responses accepted their thermal conditions. Acceptability of
thermal conditions is assumed to be equivalent to satisfaction with
thermal environment. Of the total 421 votes which found their
environment to be acceptable, 3.6% votes had a TSV of greater than
1 and 9.3% had TSV below 1. This implies that nearly 13% of the
satised votes were outside the commonly assumed comfort zone
of 1 TSV. At the same time, these gures imply cooler sensations
are more acceptable to occupants than warmer sensations.
Using the percentage of occupants who accepted their thermal
environment on each of the 19 survey days, a second order polynomial regression t is obtained between top and acceptability
votes percentage (PS). This t is given in Equation (6).
2
PS 0:70top
37:67top  409:6;

R2adj 0:69;

p < 0:001
(6)

From Equation (6), a comfort zone corresponding to 80% occupant satisfaction is found between 22.1 and 31.5
C. The plot for the
regression t and acceptability percentages at different top are also
presented in Fig. 5.
Zhang et al. [26]s analysis of ASHRAE RP-884 database revealed
that in summer time, acceptability in NV and mixed mode buildings
dropped off rapidly at 30
C on warm side and at 21.5
C on cool
side. These thresholds' are close to the limits for 80% comfort limits
found in our survey. Singh et al. [27]s survey for residences in
North India also found a similar comfort zone, i.e., between 22.5
and 30.6
C.
To examine how thermal sensation of students affects their
acceptance of the thermal environment, each day's percentage

3.1.3. Sweating sensation votes

The sweating sensation votes of individuals had a Pearson correlation coefcient (r) of 0.56 (p < 0.001) with their TSV. Similarly,
sweating sensation had the following correlation with a few other
indoor parameters:

Table 4
Comfort temperature found using Grifths equation.

Combined
2012e13 study
2013e14 study

Survey MTSV

top

tc

0.11
0.27
0.04

29.8
30.2
29.6

29.5
29.6
29.5

Fig. 5. Thermal acceptability variation with operative temperature.

402

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

acceptability (PS) was tted with the corresponding MTSV to give a

second order polynomial, Equation (7).

PS 10:34MTSV2  9:01MTSV 89:81; R2adj 0:71; p<0:001

(7)
Using Equation (7), MTSV values corresponding to 80% satisfaction were found to be 1.50 and 0.63, with a maximum
acceptability level of 91.8% at MTSV of 0.44. This result also shows
that cooler sensations are more acceptable. More importantly, this
would mean that comfort zone around neutrality is not necessarily
symmetrical. The 80% satisfaction limits and point of minimal
dissatisfaction thus obtained are close to the numbers found by
Hwang et al. [28] from their surveys done in residences and ofces
of Taiwan. Hwang et al. found 80% satisfaction limits at 1.45 and
0.65, while their point of minimal dissatisfaction was at 0.45.

3.2.1. Thermal comfort votes and thermal acceptance

Fig. 6 presents day wise thermal acceptance and thermal comfort levels during the 2013-14 survey. Similar results are not given
for the 2012-13 surveys since thermal comfort question had not
been a part of those surveys. All respondents who voted better than
Just comfortable to the thermal comfort question were deemed to
be comfortable on any given day.
On those days when acceptability is  80%, thermal comfort
levels and acceptability levels were similar. However, when
acceptability went below 80%, comfort levels deteriorated faster
and larger difference between percentages of acceptability and
comfort are seen. Day 11 was particularly bad, with a large
discrepancy between thermal acceptability and comfort. On this
day, we had recorded both the highest top and pw values.
On Day 12, even though top was 34
C, acceptability levels came
close to 80% (78%). An explanation was found in the survey responses where two students had written messages to the effect.
Due to rain the environment was comfortable today. These remarks were about the bit of unexpected rainfall that had just preceded the class. This reects on the relation between expectations
and acceptance. Since students expected a warm environment,
even the slight but unexpected, relief brought on due to rainfall
increased acceptance dramatically.

3.2.2. Thermal preference and preferred temperature

Equation (8) gives a regression relation between thermal preference vote (TP) and TSV for each individual subject.

TP 0:41TSV  0:39;

R2 0:41;

p < 0:001

(8)

This correlation shows that the thermal preference of No

change corresponds to a TSV of 0.95. As the survey location has
primarily a hotehumid climate, preference of subjects towards
cooler sensations was expected. Similar trend of thermal preference from other surveys in warm climates has been surmised in an
earlier review of comfort eld studies [10]. Percentage of occupants
who preferred warmer and cooler thermal environments on each
day were tted with logit models. The goodness-of-t of the logit
model to the observed data was assessed using a non-parametric c2
test and for both models p values were less than 0.05. Equation (9)
give the logit t models for Cooler Preference (CP) and Warmer
Preference (WP). Using these models, the preferred temperature
was found to be 26.8
C. This value is close to the preferred temperature obtained by Ogbonna and Harris [4] for university classrooms in Nigeria (26
C).

CP 0:43top  12:92

(9a)

WP 0:37top 8:55

(9b)

3.2.3. Diurnal range acceptability for temperature variation

Depending on available adaptive opportunities and clothing
adjustments across seasons, the spread of comfort zones over a year
can be quite broad. Over a single day though, available adaptive
adjustments (especially to clothing) are limited and hence the
expanse of a diurnal comfort zone should be narrower. To determine diurnal temperature variation with which occupants would
be comfortable, an analysis similar to that used by Nicol and
Humphreys [29] for European ofces was carried out. For each
survey day, using top and MTSV, a tc was calculated from Grifths'
equation. A slope of 0.5 was used for Grifths' equation. The difference between top and tc (top  tc ) was taken as comfort temperature deviation: DTc . A second order polynomial regression t
was formed between each day's percentage of occupants satised
(PS) and DTc . The resulting relation is given in Equation (10).

Fig. 6. Thermal comfort and thermal acceptability across the survey days.

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

PS 2:56DTc2  4:54DTc 89:54;

R2adj 0:71;

p < 0:001
(10)

Dtc for 80% acceptability, from Equation (10) came to [3, 1.2]
C.
Thus, occupants found diurnal temperature variation of about 4
C
to be acceptable. But larger deviations on the negative side, i.e. top
values lower than calculated tc, were agreeable. This further points
towards asymmetry of comfort zone and that in warm climates,
comfort zones have a bias towards the cooler side. The polynomial
t of Equation (10) is also presented in Fig. 7. In the plot one may
observe the sharper decline of acceptability on the positive side of
Dtc scale.
3.3. Observations regarding adaptive actions
Researchers conducting thermal comfort surveys in classrooms
have often commented upon how students often lack control upon
adaptive avenues [2,3,30e32]. This is especially true during lessons
and often opening/closing of windows or operation of fans depends
more on actions of the teacher than of the students [32]. We present in this section certain observations regarding students adaptive actions.
3.3.1. Use of fans and windows
The classroom provided occupants with standard adaptive opportunities associated with NV buildings, i.e., operable windows
with internal shades and fans. There were total of four windows,
four ceiling fans, and one wall mounted fan. Each window had two
operable panes. The ceiling fans did not have speed controllers.
Students were free to choose where they sat in the class and if they
felt like, they could also change their seat during the class.
Fig. 8 shows seating density of students across the three rows in
the classroom (Fig. 1) and usage of fan and windows on the
different survey days. Window usage percentage is calculated as
the total number of panes opened out of the available eight operable panes. Since observation of student seating was not recorded
on the rst survey day, we do not present that day's data in Fig. 8
(a). As the wall mounted fan was minimally effective compared to
the ceiling fans, usage percentage of only the ceiling fans is presented in Fig. 8 (b).
From Fig. 8 (a), a paradoxical trend is seen regarding seating
behaviour of students. They rarely chose to seat in the row next to
windows. Normally, in an NV building, it would be expected that

Fig. 7. Diurnal thermal acceptability variation with comfort temperature deviation

(DTc ).

403

occupants would want to be closer to windows. This behaviour may

be due to the distance of the particular row from the entry points
(doors). Students who entered after the class started, invariably
chose to seat in the row nearest to the doors.
Fan usage was close to 100%, except for the cooler days during
November, January, and February. Windows usage, as seen from
Fig. 8 (b), has a more complicated trend. Even on the hottest survey
day, only 50% of the operable window area was open. And this was
the only day when a student was observed opening up part of a
window during class. This action did increase the window opening
percentage to 75%. Generally, students did not pay much attention
to windows. When they entered into class, they switched on fans
and again no attention was given to state of fans through the class.
This explains why on days 8 and 9 one of the ceiling fans was not
used even though top was 30
C. The discrepancy in levels of window and fan usage may be explained by ease of operation e one
of the three E's (ease, effectiveness, and economy) [10] which occupants consider while choosing an adaptive action. Fans, due to
their ease, are more frequently used than opening windows. And
since there was no speed control available for the ceiling fans, they
always operated at peak speed once put to use. Closing/opening of
windows is mostly neglected and as observed earlier, most students do not even seat close to the windows. This may be considered as an extreme case where ease of using fans has made
occupants pay minimal attention to another adaptive action that
would require slightly more effort. Part of this may be due to occupants' lack of attention and insensitivity to their thermal environment. This could be ascribed to a high level of acclimatisation
and was also observed by Pellegrino et al. [7]. From the teacher's
side, there was per say no restriction on opening of windows or
switching fans on and off, even during the class. However, over the
12 days, only once was any such action noted during class.
Mean air velocity kept between 0.6 and 1.2 m/s on the days
when fans were in use. When fans were not switched on, air velocity was negligible. On most days though, all four fans were in use
and air velocity recorded during 12 of the 19 days (including both
2012e13 and 2013-14 surveys) was between 0.9 and 1.2 m/s. Majority (greater than 50%) responses opted for no change in air velocity on eight days while a majority voted for more air velocity on
three days. On one day, 49% wanted no change while 34% wanted
more. But on none of the days was the majority opinion for lower
air velocity. During the entire survey, 55% wanted no change in air
velocity, 31% wanted more while only 14% wanted less.
3.3.2. Clothing adaptation
As the institute did not impose any dress code, students used a
broad range of clothing ensembles with clo values between 0.31 (tshirt and light trousers) and 0.99 (full sleeve shirt and jeans with
sweatshirt). However, average clo values did not show a similar
range of variation. On the three coldest survey days, mean clo
values were between 0.62 and 0.70. On the other days, mean clo
values remained close to 0.5, the ASHRAE recommendation for
approximating summer clothing. Most students did not go for the
lowest possible clo value on warm days or the highest possible clo
value on cold days. For example, it was more common to pair up tshirt with jeans than with light trousers. This behaviour may be
explained based on what kind of ensemble is considered fashionable as well as what particular garments are available with
students.
On warm days, students often had a relaxed posture, wore light
t-shirts, and often loosened their shirt's top one/two buttons. Students who did wear full sleeves on such days, often folded up their
sleeves. Similar occupant behaviour had been observed in a couple
of other thermal comfort surveys [8,33]. Adaptive actions like
folding up of shirt sleeves or loosening a couple of buttons are

404

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

Fig. 8. Student behaviour a) seating density across rows b) window and fan usage.

primarily important in terms of giving occupants more sense of

control and exposing more skin area to direct air movement.
Table 5 gives regression R2 value between daily mean clo values
and different indices of outdoor temperature. Daily average clo
values had the strongest relation with the day's minimum temperature, which would have been recorded during early morning.
This nding was similar to that of Schiavon and Lee [34] and supports the idea that subjects' decision regarding their apparel is
greatly affected by early morning temperatures. Next to daily
minimum temperatures, average clo values have the strongest
relation with daily mean temperatures.
3.4. A comparison with ndings from the thermal comfort surveys
in undergraduate laboratory
During the Spring semester of 2013, thermal comfort surveys
were carried out in an undergraduate laboratory at IIT Kharagpur
[8]. The laboratory building is in an annex adjacent to the classroom
buildings considered in the current work. During laboratory classes,
students had a sustained metabolic rate higher than that during
classes. At the same time, laboratories have a more exible atmosphere as compared to classes. Students can move about the room,
adopt relaxed posture, and have discussions amongst themselves.
During the surveys on warm days in the laboratory, we had
observed almost every student taking multiple breaks for drinking
water. Such behaviour was not observed during classes. A comparison between the observations and ndings from both surveys
are presented in Table 6.
The outdoor and indoor conditions during both surveys were
similar though a lower top was recorded in the classrooms surveys.
While APD range was similar, MTSV range was broader for the
classrooms. This may have to do with the cooler temperature
recorded during classroom survey. The smaller range of sensations

may also be because laboratories have a three hour duration as

compared to one or two hour duration of classes. The relations that
Rohles and Nevins developed from their study on college age students [14] indicate that as time of exposure increases, magnitude of
thermal sensation decreases.
Both regression neutral temperature and preferred temperature
were lower for the laboratory studies and by about an equal
magnitude. This may be ascribed to the higher metabolic rates in
lab activity. The mean comfort temperature (tc) calculated from
Grifths equation is also lower for the lab study though not as
much lower as the regression neutral temperature. The regression
equation slope is smaller for the lab study and this may suggest that
there is more complete expression of adaptive action in those
studies. However, it may also be because the laboratory studies
encountered as broad a temperature range as the classroom studies
but over a single semester instead of two.
More interesting are the comfort zones from the two studies.
The lower comfort zone limit from laboratory study carries less
import as temperatures below 22
C were not encountered. The
upper limits of both comfort zones are nearly equal though (difference of 0.6
C). This would suggest that the higher metabolic rate
in laboratories was about effectively negated by availability of more
adaptive opportunities. Breadth of diurnal comfort band was about
4
C for both studies. But while the band was symmetric for lab
studies, it was biased towards cooler deviations for the classroom
studies. This may again be due to more adaptive opportunities in
laboratory that allow subjects to deal with positive deviations
equally well as negative deviations between top and tc.
Fig. 9 tries to present a summation of all 31 days of thermal
comfort survey (12 from laboratory and 19 from classrooms) in
terms of temperature and humidity combinations that were
acceptable to students. From 21 to 30
C, conditions recorded were
always acceptable to more than 80% of occupants. Over 30
C,

Table 5
Correlations between outdoor indices and mean clo values.

R2/R2adj

Daily mean

RMT

Daily min

Daily max

PMOAT

MMT

Last day's mean

0.70/0.67

0.61/0.57

0.76/0.74

0.50/0.45

0.59/0.55

0.56/0.51

0.69/0.66

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

405

Table 6
Comparison between studies from laboratory and classroom.

Outdoor daily mean

top
MTSV
APD
Regression equation
Regression neutral temperature
tc (Grifths' equation)
Preferred temperature
Comfort zone (from 2nd order polynomial t of top and % satisfaction)
Diurnal comfort zone width

Fig. 9. 80% thermal acceptance spread over operative temperature and humidity.

higher humidity values, e.g. pw > 2:6 kPa, become unacceptable.

Above 31
C, conditions are mostly unacceptable except for one or
two exceptions.
Of the 31 days on which thermal comfort surveys were conducted, acceptability of thermal conditions were less than 80% on
10 days. Analysis of these 10 days showed that except for an early
morning (7:30 am) class in January, the other nine days were during
April and March. Of these nine cases, 7 were during afternoon. This
shows a strong indication that March and April are the problem
months with the afternoon hours being of particular concern.
While this may not be surprising eee as March and April are the
warmest months in Kharagpur during which regular classes take
place ee this trend does suggest a simple precaution. As IIT has a
residential campus, classes during March and April may be
scheduled during early morning/late evening hours so as to limit
student discomfort. Similarly, for January, which is the coldest
month, classes may be started later in the day, say 8:30 am instead
of 7:30 am, to avoid the most uncomfortable early morning hours.

4. Conclusion
This study presented results from thermal comfort surveys
conducted across an academic year in NV classrooms, in a tropical
region. A 9
C broad comfort zone is determined while the neutral
temperature found is close to 29
C. These results clearly support
the use of adaptive comfort standards for NV classrooms in tropics.
The students were well acclimatized to the local climate. They

Laboratory

Classroom

33e14.5
C
35e22
C
1.74e(0.73)
59%e0%
TSV 0:184top  4:866
26.4
C
28.2
C
24
C
19.4e30.9
C
2

32.5e14.5
C
35e20
C
2e(1.8)
56.8%e0%
TSV 0:22top  6:367
29
C
29.5
C
26.7
C
22.1e31.5
C
3; 1:2

displayed adaptive behaviours like use of fans, clothing adjustment,

and window operation. Analysis shows that diurnal temperature
variation of about 4
C will be tolerable to nearly 80% occupants. But
the results also show a trend of subjects preferring cooler thermal
sensation and deviations on the cooler side of neutral being more
tolerable. Comparison with the comfort studies done in laboratory,
at the same location, shows that both studies have similar upper
limit for their comfort zones. This leads us to believe that greater
adaptive opportunities in a laboratory setting e as compared to
classroom e can offset the higher metabolic rates prevalent for
laboratory level activity. Additionally, combined results from both
studies show that the buildings investigated provided acceptable
thermal conditions to occupants over most of the operational
period. With some adjustment to scheduling, these classrooms can
ensure student comfort and are unlikely to need intervention of
mechanical conditioning.
The ndings from this study would be helpful for developing
futuristic comfort standards for NV classrooms in India. As a fast
developing nation, number of classrooms in India is also rising
quickly. Our results show that with appropriate avenues of adaptation, comfort in NV classrooms is achievable. Thus adaptive
comfort standards can be the guiding principle for the future of
energy efcient and comfortable classrooms in Indian context.
Acknowledgements
We are grateful for the cooperation from all subjects of these
thermal comfort studies. We also appreciate the help extended by
Mr. Ashish Anand and Mr. B. Dinesh Reddy, undergraduate students, in collecting the data during 2012e13 surveys.
References
[1] Planning Commission Report, Annual Status of Higher Education of States and
UTs in India, (Summary Report), Planning Commission. Government of India,
New Delhi, India, 2012.
[2] A.G. Kwok, Thermal comfort in tropical classrooms, ASHRAE Tran. 104 (1)
(1998) 1031e1050.
[3] N.H. Wong, S.S. Khoo, Thermal comfort in classrooms in the tropics, Energy
Build. 35 (4) (2003) 337e351.
[4] A.C. Ogbonna, D.J. Harris, Thermal comfort in sub-Saharan Africa: eld study
report in Jos-Nigeria, Appl. Energy 85 (1) (2008) 1e11.
[5] I. Hussein, M.H.A. Rahman, T. Maria, Field studies on thermal comfort of airconditioned and non air-conditioned buildings in Malaysia, in: Energy and
Environment, 2009. ICEE 2009. 3rd International Conference on. IEEE, 2009,
pp. 360e368.
[6] C. C^
andido, R. de Dear, R. Lamberts, Combined thermal acceptability and air
movement assessments in a hot humid climate, Build. Environ. 46 (2) (2011)
379e385.
[7] M. Pellegrino, M. Simonetti, L. Fournier, A eld survey in Calcutta. Architectural issues, thermal comfort and adaptive mechanisms in hot humid climates,
in: Proceedings of 7th Windsor Conference: The Changing Context of Comfort
in an Unpredictable World, NCEUB, Windsor, UK, 2012.
[8] A.K. Mishra, M. Ramgopal, Thermal comfort in undergraduate laboratories e A
eld study in Kharagpur, India, Build. Environ. 71 (2014a) 223e232.
[9] A.K. Mishra, M. Ramgopal, Thermal comfort eld study in undergraduate

406

[10]
[11]
[12]

[13]
[14]
[15]
[16]

[17]
[18]

[19]

[20]
[21]
[22]

[23]

A.K. Mishra, M. Ramgopal / Building and Environment 92 (2015) 396e406

laboratories e An analysis of occupant perceptions, Build. Environ. 76 (2014b)
62e72.
A.K. Mishra, M. Ramgopal, Field studies on human thermal comforte An
overview, Build. Environ. 64 (2013) 94e106.
D. Teli, P.A. James, M.F. Jentsch, Thermal comfort in naturally ventilated primary school classrooms, Build. Res. Inf. 41 (3) (2013) 301e316.
A.K. Mishra, M. Ramgopal, Thermal comfort in classrooms in tropics: an
analysis of student preference, in: Proceedings of Conference Efcient, High
Performance Buildings For Developing Economies, 2014, ASHRAE, April
2014c.
Vaisala Humidity Calculator for Windows, Version 3.0, 2013. URL, www.
vaisala.com/humiditycalculator.
ASHRAE, 2009 ASHRAE Handbook: Fundamentals, SI ed., American Society of
Heating, Refrigerating and Air-conditioning Engineers, Atlanta, GA, 2009.
J.F. Nicol, I.A. Raja, A. Allaudin, G.N. Jamy, Climatic variations in comfortable
temperatures: the Pakistan projects, Energy Build. 30 (3) (1999) 261e279.
F. Nicol, M. Humphreys, Derivation of the adaptive equations for thermal
comfort in free-running buildings in European standard EN15251, Build. Environ. 45 (1) (2010) 11e17.
SI/ASHRAE, Standard 55e2013. Thermal Comfort Conditions For Human Occupancy, ASHRAE, Atlanta, 2013.
M. Humphreys, H. Rijal, J. Nicol, Updating the adaptive relation between
climate and comfort indoors; new insights and an extended database, Build.
Environ. 63 (2013) 40e55.
R Core Team, R: A Language and Environment for Statistical Computing, R
Foundation for Statistical Computing, Vienna, Austria, 2014. URL, http://www.
R-project.org/.
D.A. Kenny, Statistics for the Social and Behavioral Sciences, Little Brown,
Boston, 1987.
F.H. Mallick, Thermal comfort and building design in the tropical climates,
Energy Build. 23 (3) (1996) 161e167.
M. Indraganti, Using the adaptive model of thermal comfort for obtaining
indoor neutral temperature: ndings from a eld study in Hyderabad, India,
Build. Environ. 45 (3) (2010) 519e536.
H.H. Liang, T.P. Lin, R.L. Hwang, Linking occupants' thermal perception and

[24]

[25]
[26]
[27]

[28]

[29]
[30]

[31]
[32]

[33]

[34]

building thermal performance in naturally ventilated school buildings, Appl.

Energy 94 (2012) 355e363.
C.P. Chen, R.L. Hwang, W. Liu, W.M. Shih, S.Y. Chang, The inuence of airconditioning managerial scheme in hybrid-ventilated classrooms on students' thermal perception, Indoor Built. Environ. (2014), http://dx.doi.org/
10.1177/1420326X14530587.
F. Nicol, Adaptive thermal comfort standards in the hotehumid tropics, Energy Build. 36 (7) (2004) 628e637.
H. Zhang, E. Arens, W. Pasut, Air temperature thresholds for indoor comfort
and perceived air quality, Build. Res. Inf. 39 (2) (2011) 134e144.
S. Singh, P. Chani, S. Kulkarni, Implication of building energy modeling (BEM)
and adaptive model to assess the efciency of multi storied apartments in
composite climate of north India, in: Proceedings of 8th Windsor Conference:
Counting the Cost of Comfort in a Changing World, NCEUB, Windsor, UK,
2014.
R.L. Hwang, M.J. Cheng, T.P. Lin, M.C. Ho, Thermal perceptions, general
adaptation methods and occupant's idea about the trade-off between thermal
comfort and energy saving in hotehumid regions, Build. Environ. 44 (6)
(2009) 1128e1134.
F. Nicol, M. Humphreys, Maximum temperatures in European ofce buildings
to avoid heat discomfort, Sol. Energy 81 (3) (2007) 295e304.
S.P. Corgnati, R. Ansaldi, M. Filippi, Thermal comfort in Italian classrooms
under free running conditions during mid seasons: assessment through
objective and subjective approaches, Build. Environ. 44 (4) (2009) 785e792.
A.G. Kwok, C. Chun, Thermal comfort in Japanese schools, Sol. Energy 74 (3)
(2003) 245e252.
L. Dias Pereira, D. Raimondo, S.P. Corgnati, M. Gameiro da Silva, Assessment of
indoor air quality and thermal comfort in Portuguese secondary classrooms:
methodology and results, Build. Environ. 81 (2014) 69e80.
M. Indraganti, R. Ooka, H.B. Rijal, Thermal comfort in ofces in summer:
ndings from a eld study under the setsuden conditions in tokyo, japan,
Build. Environ. 61 (2013) 114e132.
S. Schiavon, K.H. Lee, Dynamic predictive clothing insulation models based on
outdoor air and indoor operative temperatures, Build. Environ. 59 (2013)
250e260.