Building and Environment 188 (2021) 107428

Contents lists available at ScienceDirect

Building and Environment

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

Unlocking thermal comfort in transitional spaces: A field study in three

Italian shopping centres
Marta Avantaggiato a, Annamaria Belleri a, Ulrich Filippi Oberegger a, Wilmer Pasut b, *
a
Eurac Research, Institute for Renewable Energy, Bolzano, Italy
b
Universita’ Ca’ Foscari, Venezia, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Shopping centres are commonly laid out as small individual stores connected by transitional spaces. Setpoint
Thermal comfort temperatures used to control transitional spaces are normally the same as in traditional indoor environments
Transitional spaces despite substantial differences in use, time of permanence and users’ needs. Currently, there are no comfort
Shopping centre
guidelines for transitional spaces and the literature lacks relevant studies on the topic. There is an untapped
Adaptability
Indoor environment quality
potential for energy savings and improved indoor environmental quality. The main objective of this work is to
evaluate the suitability of Fanger’s comfort model and adaptive comfort model for transitional spaces. We
assessed users’ thermal perception and potential impacting factors in three Italian shopping centres. 724 cus­
tomers were interviewed on their thermal comfort, thermal sensation, thermal preference, and clothing level
while experiencing the transitional space. In addition, the thermal environment at the interview locations (dry-
bulb temperature, globe temperature, relative humidity, and air speed at different levels) and the outdoor
temperature were monitored. The study demonstrated that Fanger’s model and the adaptive comfort model are
not suitable for transitional spaces. Customers were inclined to adapt to a much wider range of indoor envi­
ronmental conditions. An operative temperature of up to 27.5 ◦ C was still deemed comfortable by more than 80%
of the customers. These results unlock a large potential for energy savings and pave the way for passive solutions
such as natural ventilation.

1. Introduction individual system per each retail unit. In either case, the temperature
setpoints inside the single shops are controlled by managers, workers, or
Large shopping centres are based on a model of small individual according to brand policies. Most of the time this results in temperatures
stores connected by common areas that enable customers to move from a that are different from the common areas, which exposes customers to
shop to another without exiting the centre. They can be shaped as gal­ several rapid changes in thermal conditions.
lery, atrium or ring and can be located on a single level or connect Moreover, indoor temperatures are generally controlled indepen­
multiple levels. dently from outdoor conditions. Implemented setpoints are typically
The number of people in the common areas varies during the day. based on guidelines intended for other indoor environments or on the
Observed users’ activity is dynamic and the time of permanence differs, experience of the facility manager. The customers’ clothing and their
which creates highly variable occupancy. activity are not (or cannot be) taken into consideration and sometimes
Shopping centres are normally conditioned by an all-air HVAC sys­ results in uncomfortable conditions or inefficient energy management.
tem that handles both the individual stores and retail units or using an Due to their unique features, common areas may not require the same

Abbreviations: MEMO, Mobile Environmental Monitoring; PMV, Predicted Mean Vote; PPD, Predicted Percentage of Dissatisfied; TAV, thermal acceptability vote;
TCV, thermal comfort vote; TPV, thermal preference vote.
* Corresponding author. Scientific Campus, via Torino, 155, 30170, Venice, Italy.
E-mail address: [email protected] (W. Pasut).

https://doi.org/10.1016/j.buildenv.2020.107428
Received 5 August 2020; Received in revised form 16 October 2020; Accepted 31 October 2020
Available online 4 November 2020
0360-1323/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

tight control as other indoor environments. A wider range of setpoints whereas transitional spaces have dynamic features making Fanger’s
may still provide or even improve comfort. model unsuitable. Some studies [8–11] also observed that users in
Thermal comfort research traditionally focuses on either the indoor transitional spaces have a higher adaptability, which might be influ­
or the outdoor environment. Little has been discussed in previous work enced by the time spent within the transitional space and the previous
about people’s thermal perception and comfort in indoor transitional thermal experience. All these studies concluded that further investiga­
spaces1 with transient and dynamic conditions. Table 1 collects studies tion on thermal comfort in transitional spaces at different times of the
related to thermal comfort in transitional spaces in the last fifteen years. year is required to expand the database of evidences.
Only field studies have been included in the table because they are Current comfort standards do not specifically address transitional
representative of the specific conditions in the transitional spaces. spaces, which are therefore treated by energy managers like other in­
However, we briefly report on important findings from studies con­ door environments (e.g. offices) [6]. According to Pitts and al. [4], the
ducted in more strictly controlled environments such as labs and climate energy demand of transitional spaces per unit area or volume is three
chambers. Zhang et al. [1] analysed psychological and physiological times that of the remaining of the building interior, and past studies [4,
responses from thirty college students subjected to step changes of 7–10] identified a high energy use intensity of transitional spaces for
neutral-warm and neutral-cool in a climate chamber consisting of two heating and cooling and thus a great energy saving potential linked to
adjacent rooms in China reproducing a hot-humid climate. As accept­ more relaxed setpoint settings.
able upper limits for transitional spaces and people naturally acclima­ The main objective of this work is to evaluate the suitability of
tized to such a climate, air temperatures of 29.2 ◦ C at 50% relative Fanger’s model and adaptive comfort model for transitional spaces. We
humidity and 28.0 ◦ C at 70% relative humidity for 90% satisfied were also investigated the interrelationship of users’ thermal acceptability,
reported. Despite the challenge to represent natural movement of people comfort, and preference as well as the range of operative temperatures
in transitional spaces within such a controlled environment, this study considered comfortable by most users.
shows the high adaptation potential of people accustomed to a specific
climate. 2. Methods
The common methodological approach is to combine objective
measurements with subjective questionnaires. Most of the studies were We conducted a field study in three Italian shopping centres in spring
conducted in the UK [4,8,10,11,13] and in the south-east of Asia [2,5–7, and summer 2016. Detailed outdoor and indoor environmental moni­
12,14–16]. No studies were found concerning transitional spaces in toring was performed while the customers were asked to complete a
Southern Europe except for three papers focusing on the same hyper­ dedicated questionnaire on their perception over the indoor parameters.
market in Bari, Italy. In the first paper [17], the authors perform physical 724 customers answered questions about thermal comfort, thermal
measurements as well as administer IEQ questionnaires to over 300 sensation, thermal preference, and clothing while being in a common
workers in the hypermarket. They found that Fanger’s model [20] was area of the shopping centre.
less suitable to describe the thermal sensation in the naturally ventilated
warehouses than in air-conditioned areas. Therefore, they suggested to 2.1. Environmental monitoring
investigate more suitable indexes for these environments. In the second
paper by the same authors [18], they included further data collected The measurement campaign considers the monitoring of both indoor
after their earlier paper and investigated the influence of clothing dis­ and outdoor environmental variables as described in the following
tribution and local discomfort on the Fanger model’s accuracy. As these sections.
studies focused on specific aspects, no upper limits for acceptable
thermal conditions in such spaces during the hot season were reported. 2.1.1. Indoor environmental monitoring
In the third paper [19], statistical techniques were applied to the data to For this study, a cart named MEMO (Mobile Environmental MOni­
derive, through a multiple regression, ‘optimal ranges’ for thermal toring) was made in the Eurac Research labs. MEMO can be easily
conditions considered satisfactory by the workers, namely 13.7–21.8 ◦ C moved on a flat floor and the height of the sensors can be modified as
(17.2–22.8 ◦ C) in winter (summer) for operative temperature and needed. MEMO was equipped with a globe thermometer and hygrom­
16.4–30.3 ◦ C for floor temperature. The low operative temperature eter to respectively measure globe temperature and relative humidity at
range for summer derived from the regression was lower than the 1.1 m above the floor. Sensors were placed at 1.1 m and 1.6 m to monitor
measured values and may be due to the workers’ activities, clothing, and air speed and dry bulb temperature (see Fig. 1).
individual preferences, and possibly the statistical method used, among The accuracy of the sensors meets the recommendations of the Eu­
potential other factors. Another paper dealing with transitional spaces in ropean standard EN ISO 7726 [23]. Air temperature is measured using
Southern Europe is the one of Albuquerque et al. [21]. I this case the radiation shielded Pt100 sensors. To measure the globe temperature, a
focus was more on control strategies during closing time and related 40 mm globe thermometer was built from a ping-pong ball painted in
energy savings more than users’ thermal comfort. grey on the inside and opaque black on the outside. This setup closely
All above studies refer to Fanger’s model and do not consider the approximates the operative temperature for limited air speed for indoor
adaptive thermal comfort model [22]. applications [24,25]. The mean radiant temperature was determined
Most of the studies [2–4,7,8,11,14,16,12] concluded that people in from the globe thermometer measurement as per European standard EN
transitional spaces can accept a wider range of comfort conditions than ISO 7726 [23]. Omnidirectional hot wire sensors were used to measure
predicted by Fanger’s model. In their view, the limitation of Fanger’s air speeds at the two heights. Indoor relative humidity was measured
model lies in the steady-state conditions under which it was developed, with a portable probe. All thermal parameters where measured and
recorded every 10 s. Table 2 presents the characteristics of the used
sensors.
1
Indoor measurements were performed in several locations within the
Referring to shopping centres, transitional spaces are all those conditioned
building, mainly atriums and galleries between more frequented shops.
areas open to the public that are within the building but are not the shops. It is
Areas with direct sunlight were limited in the three case studies and
the connection area between the different shops, which is made by elements
like atriums, corridors, entrance, where customers can pass through or rest or anyway never used for the measurements.
do other activities while in the shopping centre. As expressed by Pitts et al. [4],
these spaces are often perceived as some of the most important in architectural 2.1.2. Outdoor environmental monitoring
design terms since they also impact on a wide range of senses and perceptions of Outdoor dry bulb air temperature and relative humidity were
human occupants. measured every 10 s. To this aim, we used a MEMS (Micro-Electro-

2
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

Table 1
Field studies related to transitional spaces.
Author Building type Research method Location Year Ref.

Jitkhajornwanich et al. Educational building Questionnaire Bangkok, Thailand 2002 [2]

Office Physical measurements
Chun et al. Lobbies, balconies, pavilions Physical measurements (long & short term) Yokohama, Japan 2004 [3]
Observation (activities)
Pitts et al. Educational buildings Questionnaire Sheffield, UK 2008 [4]
Physical measurements
Hwang et al. Entrance Questionnaire Taichung, Taiwan 2008 [5]
Atrium Physical measurements
Service centre
Kwong et al. Lobby Questionnaire Serdang, Malaysia 2009 [6]
Educational building Physical measurements
CFD simulations
Hui and Jie Lift, lobbies, corridors Educational Questionnaire Hong Kong 2014 [7]
building Physical measurements Energy simulation
tool
Kotopouleas A., Airport terminal Questionnaire Manchester London, 2016 [8]
Nikolopoulou Physical measurements UK
Mishra et al. Museum Questionnaire Amsterdam, 2016 [9]
Physical measurements Netherlands
Vargas Lobby Questionnaire Sheffield, UK 2016 [10]
Physical measurements
Hou Atria, educational buildings, business Questionnaire Cardiff, UK 2016 [11]
centre Physical measurements
Li et al. Underground malls Questionnaire Nanjing, China 2018 [12]
Physical measurements
Tse et al. Shopping centre Questionnaire Cardiff, UK 2019 [13]
Physical measurements
Du et al. Shopping mall Questionnaire Beijing, China 2020 [14]
Physical measurements
Kwok et al. Shopping centre Questionnaire Hong Kong 2017 [15]
Pin Lu, Jin Li Commercial building Questionnaire Guangzhou, China 2020 [16]
Physical measurements
Martellotta et al. Supermarket Questionnaire Bari, Italy 2012-2013- [17–19]
Physical measurements 2016

Fig. 1. Mobile environmental monitoring cart “MEMO” and parameters measured at different levels.

Mechanical Systems) integrated portable data logger whose features are participate in an interview. We filled their answers in a questionnaire
listed in Table 3. Measurements were performed in outdoor parking structured into three main sections: background, thermal comfort, and
areas of the three shopping centres. clothing.

2.2. Interviews • Background - general questions about age, gender, health condi­
tions, if they had eaten or drunk, and their previous activities.
Customers passing by our equipment were randomly asked to

3
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

Table 2 shopping centres and climates according to the Köppen-Geiger classifi­

Measured indoor environment parameters and sensor characteristics. cation [28] are presented in Table 4.
Parameter Sensor Measuring Accuracy The first two centres (SC01 and SC02) are in the municipality of
range Trento in the north of Italy and the third centre (SC03) is in Catania in
Air temperature Pt100 class A − 50 to ±0.2 ◦ C (− 25 the south of Italy. Trento has a temperate, fully humid climate with hot
Radiation-shielded 150 ◦ C to 74.9 ◦ C) summers (Cfa) whereas the climate of Catania is Mediterranean with
Mean radiant Pt100 class A and 40 − 50 to ±0.2 ◦ C (− 25.0 dry, hot summers (Csa).
temperature mm diameter globe 150 ◦ C to 74.9 ◦ C) The first measurement campaign was conducted in SC1 on April 4th-
Air speed Hotwire 0.05–5.0 m/ 0.02 m/s +
omnidirectional s 1.5% of reading
6th and June 10th, 2016. The shopping centre was built in 2000 and has
anemometer a floor area of 9′ 774 m2 laid out over two floors with a total of 55 retail
Relative humidity EE08 series HC101 0–100% ±2% RH units. The common areas are mainly shop galleries. The main entrance
RH% (at ambient sensor (0–90% RH) atrium has a fully glazed, south-west-oriented façade coated with a sun
pressure) ±3% RH
control film. The field study was performed in different locations within
(90–100% RH)
the common areas including shop galleries and atria on the ground and
first floor.
The second measurement campaign in SC02 was performed on June
Table 3
21st - 22nd, 2016. The total floor area is laid over three floors with a
Measured outdoor environment parameters and characteristics of the MEMS
total of 47 retail units. The common areas are mainly shop galleries. The
integrated portable data logger.
main entrance is an atrium with a fully glazed, south-east oriented
Parameter Sensor and brand type Measuring Accuracy façade. The field study was performed in different locations within the
range
shop gallery on the first floor.
Outdoor MEMS integrated portable − 30.0 to ±0.5 ◦ C The last measurements were conducted in Catania on six days be­
temperature data logger 70.0 ◦ C
tween July 13th and 20th, 2016. SC3 was built in 2009 and contains a
Outdoor relative MEMS Integrated portable 0–100% ±2%
humidity data logger two-storey gallery with over 60 retail units, offering a gross leasable area
of 27′ 521 m2 of which 8′ 000 m2 are dedicated to a hypermarket. The
main building façade is oriented towards south-east. The field study was
Customers were also asked about the duration of their stay inside the performed in different locations within the shop gallery at the second
building before taking the survey. floor.
• Thermal comfort - customers were asked about their vote on ther­ Before the measurement campaigns we had a one-day pilot study in
mal acceptability (TAV), sensation (TSV), preference (TPV), and August 2015 in SC1 to configure the monitoring devices, test the survey
comfort (TCV) while spending time in the common area. Thermal and refine the interview process.
acceptability was assessed on a 2-point scale (acceptable or not
acceptable). For thermal sensation, a 7-point scale was used ac­
2.4. Data analysis
cording to PMV (Predicted Mean Vote) from − 3 to +3 corresponding
to the categories “cold,” “cool,” “slightly cool,” “neutral,” “slightly
A total of 724 randomly selected customers where interviewed on a
warm,” “warm,” and “hot”. Thermal preference was surveyed using a
voluntary basis during the three measurement campaigns. After a brief
3-point scale: “right now I want the environment to be: cooler, no
explanation about the content of the study, they could decide to be part
change, warmer”. Thermal comfort was evaluated on a 6-point scale
of it or not. All interviews were included in the sample, also those
(very comfortable, comfortable, just comfortable, just uncomfort­
resulting from customers who have eaten or drunk in the previous 20
able, uncomfortable, and very uncomfortable) [26].
min before the questionnaire. This to have a better representation of
• Clothing - we asked the interviewees to report their clothing to es­
typical shopping centre customers, who are used to drink coffee or eat
timate the thermal resistance according to the EN ISO 7730 standard
ice cream before or while shopping.
[27]. Physiological parameters were not measured.
The survey results were compared against Fanger’s model [20] and
the adaptive comfort model. Fanger’s model was considered because
2.3. Case studies transitional spaces are typically mechanically conditioned. However,
due to their peculiar conditions and use we decided to also evaluate the
The measurement campaigns were conducted in the transitional adaptive model suitable for free-running or naturally ventilated build­
spaces of three Italian shopping centres. The main features of the ings. Specifically, we refer to the method presented by Nicol and

Table 4
Characteristics of the three shopping centres.
Shopping Typology [29] Municipality Size Number of Climate Floors Average time of Map
centre [m2] shops permanence

SC01 neighbourhood Trento 9774 55 Cfa 2 25min - 1 h

centre
SC02 community centre Trento 6898 40 Cfa 3 20min - 1 h
SC03 super regional centre Catania 27,521 70 Csa 2 25min - 1.5 h

4
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

Humphreys [30] and included in the European standard EN 16798–1: Table 5

2019 [31]. Indoor and outdoor environmental conditions for the three case studies. For each
Using data collected by the monitoring cart (MEMO) and the clothing shopping mall, Mean, SD, Min, and Max denote the arithmetic mean, standard
ensembles recorded during the interview, it was possible to calculate deviation, minimum value, and maximum value, respectively, of all measured
PMV and PPD (Percentage of People Dissatisfied) for the indoor condi­ values.
tions experienced by the customers during the time they took the survey. Indoor Outdoor Average
The metabolic activity was assumed equal to 1.6 met (shopping) for all clothing
level
customers [27]. The R package “comf” [32] was used for the PMV-PPD
Toperative V T
calculation. air RH dry bulb RH Clo
(◦ C) (m/s) (%) (◦ C) (%)
Because of the nature of the seven-point scale used to assess cus­
tomers’ thermal sensation, the TSV is a categorical variable, which is not SC01 Mean 25.1 0.12 48 24.3 43 0.64
directly comparable with a continuous variable such as the PMV. A SD 0.9 0.06 7 2.1 7 0.2
Min 22.9 0.00 59 19.4 31 0.18
direct comparison between TSV and PMV is however commonly done Max 26.7 0.33 33 29.5 55 1.41
with a binning of the PMV by setting all values lower or equal − 2.5 to SC02 Mean 25.3 0.12 48 30.8 40 0.45
− 3, higher than − 2.5 and lower or equal − 1.5 to − 2, and so on. This SD 0.5 0.08 2.9 1.3 2 0.1
binning was realized with the function “cutTSV” of the R package Min 24.4 0.00 43 28.5 35 0.23
Max 26.3 0.40 53 32.6 46 0.85
“comf” [32].
SC03 Mean 26.1 0.14 42 30.1 42 0.38
Since we recorded outdoor conditions only during the measurement SD 1.1 0.08 3 1.7 14 0.1
campaigns, data coming from weather stations located closer to the Min 21.7 0.00 33 27.4 17 0.22
three case studies [33] were used to calculate the mean running Max 29.6 0.41 49 34.6 62 0.67
temperature.
The evaluations of the thermal environment using the two thermal
comfort models were compared with the real customers’ satisfaction
votes coming from the surveys.
In a second phase, customers perception of the thermal environment consider the whole sample regardless of the time of permanence and the
was deeply investigated by analysing their answers in connection with activities they were carrying on 20min before the interview.
experienced indoor and outdoor parameters.
3.1.1. Indoor conditions
3. Results Table 5 shows basic descriptive statistics about the indoor and out­
door conditions in the three case studies during the measurements, as
3.1. Descriptive statistics well as the average clothing level of the customers’ interviewed. The
operative temperatures recorded during the measurement campaigns
We were able to obtain an even distribution across case studies in ranged between 21.7 ◦ C and 29.6 ◦ C. Both extreme values of the range
terms of gender and age. 59% of the interviewees were female, 44% were recorded in SC03. Indoor relative humidity ranged between 33%
were less than 30 years old, 33% were between 31 and 50 years old, and and 59%. The air speed measurements generally showed very limited
23% were older than 50. values typical of mechanically ventilated buildings. Outdoor dry bulb air
Fig. 2 shows the distribution of the time of permanence by case study temperature during interviews ranged between 19.4 ◦ C and 34.6 ◦ C. The
as indicated during the interviews. Looking at the total sample, 33.7% of lowest value was measured in SC01 during the mid-season campaign
the interviewees indicated less than 10 min of permanence, 26.8% had (April 2016) while the highest was recorded in SC03 in July 2016.
spent a period between 10 and 20 min, and 39.5% were inside the In SC01, the measurements took place in April and June. This jus­
shopping centre for more than 20 min. tifies an average value of clothing higher than for SC02 and SC03 where
39% of respondents had drunk and 21% of respondents have eaten in the measurements were performed only in summer. The clothing level
the previous 20min. On the attempt to represent the entire population of was often lower than 0.50 clo, which is the value used to estimate
shopping centre customers, in the following analysis we decided to comfort temperatures under summer conditions [27].

Fig. 2. Time of permanence within the three case studies and in total.

5
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

Fig. 3. Frequency distribution of the indoor parameters measured during the three campaigns: air temperature, operative temperature, indoor air velocity and
relative humidity.

Fig. 4. Frequency distribution of the outdoor parameters measured during the three campaigns: dry-bulb temperature and relative humidity.

The data collected in the three shopping centres were aggregated to

investigate the frequency distribution of the main environmental pa­
rameters. Fig. 3 reports the frequency distributions of the indoor pa­
rameters, while Fig. 4 shows the ones of the outdoor parameters.

The highest frequency of outdoor dry-bulb temperature distribution

Fig. 5. Frequency distribution of the difference between outdoor and indoor
ranges between 28 ◦ C and 31 ◦ C (around 45% of the time). The indoor temperature during the three campaigns.
air temperature frequency distribution is centred between 24.5 ◦ C and
25.5 ◦ C, accounting for around 45% of the total. The highest frequency
conditions. This is the cause of the negative values of the outdoor-indoor
of the operative temperature distribution ranged between 25 ◦ C and
temperature difference visible in Fig. 5.
26 ◦ C, accounting for around 47% of the total. Most of the measured
indoor air velocities were lower than 0.25 m/s. Measured indoor relative
humidity is normally distributed with peak between 40% and 50%. 3.2. Investigation of customers’ perception of the thermal environment
Finally, the difference among the outdoor relative humidity distribution
was relatively small, with the highest frequency ranging from 35% to In the following sections, the distribution of the customers’ answers
45%. related to thermal acceptability (TAV), comfort (TCV), and preference
The distribution of indoor air temperatures is slightly different from (TPV) are presented and discussed.
the distribution of the operative temperatures. This is due to the radiant Fig. 6 shows the interrelationship between thermal acceptability,
effect of internal gains such as lights and the effect of solar radiation comfort, and preference. Data related to a TSV equal to − 3, − 2, or 3
passing through the wide glazed façades that characterize shopping were removed because the number of respondents per each such TSV
centre transitional spaces. was less than 10.
The measurements were performed in both mid-season and summer Focusing first on the number of interviewed people, most people

6
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

Fig. 6. Per each thermal sensation vote it is presented “Acceptability”, first bar, “Comfort”, second bar, and “Preference”, third bar. On top, the number of responses
per each TSV are reported.

reported a TSV between − 1 (slightly cool) and 1 (slightly warm). More than the ensemble of “comfortable” and “very comfortable” conditions,
than 50 people reported a TSV of 2 (warm). Only very few people re­ and comfort is less difficult to achieve than a “no change” preference.
ported TSVs of − 3, − 2, or 3. Most people considered the environment The target “just comfortable” or a higher level of comfort is achieved in
thermally acceptable and comfortable. Nevertheless, when asked for more cases than thermal acceptability, but the difference in the number
their thermal preference, most people indicating a TSV of − 1 or 0 stated of people between “acceptable” and “just comfortable” or a higher level
that they would prefer a cooler environment. Expectedly, some people of comfort is rather small, i.e., only very few people might state that the
indicating a TSV of − 1 or 1 stated that they would prefer a warmer or environment is “unacceptable” but nevertheless “just comfortable”.
cooler environment, respectively. The strong link between acceptability, comfort, preference, and
sensation also indicates a high coherence in the customers’ answers.
To investigate the tolerance to a range of operative temperatures,
thermal sensation votes were grouped into three categories.

• TSV (− 1,0,1) stands for customers that are satisfied with the thermal
environment.
• TSV (− 3,-2) stands for customers that are cold dissatisfied
Looking at the percentages, the strong link between thermal • TSV (+3,+2) stands for customers that are warm dissatisfied.
acceptability, comfort, and preference becomes clear, but important
differences are observed as well. Acceptance is less difficult to achieve By grouping the thermal sensation votes in this way, for each

Fig. 7. Bar chart of the distribution of thermal sensation over operative temperature intervals [◦ C].

7
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

operative temperature interval, it was possible to identify the reason Table 6

that generates discomfort (too cold or too warm). Are the results in this study on thermal comfort of customers in shopping malls
A general satisfaction with the environment is observed, from 23 ◦ C generalisable?
up to 28 ◦ C (Fig. 7). A great tolerance to a wide range of operative Operative temperature Number Estimate Is the result
temperature is hence demonstrated. During only three interviews the interval midpoint (◦ C) (n) for p generalisable?
operative temperature was higher than 28.5 ◦ C and the environment was 22 5 1.00 No
perceived as too warm. 23 11 0.73 No
24 48 0.94 No
25 247 0.90 Yes***
3.2.1. Customers thermal comfort
26 284 0.84 Yesa
The customers were asked to rate their general state of comfort using 27 104 0.92 Yes***
a six-point scale. The results are reported in the bar chart in Fig. 8. 28 21 1.00 No
For most operative temperature intervals, people expressed their 29 3 0.66 No
vote at the “just comfortable” to “very comfortable” side of the scale. a
95% (***99.9%) confidence level.
The results presented in Fig. 8 suggest that customers judge as
comfortable a wide range of indoor operative temperatures. This result
interval for p is above 80%.
is in line with previous findings about acceptable temperatures in
According to these results (Table 6), an operative temperature of up
transitional spaces [2–4,7,8,11].
to 27.5 ◦ C in summer is still deemed comfortable by at least 80% of the
customers of a shopping centre that fits into the context of this study
3.2.2. Generalizability of results
regarding parameters such as outdoor and indoor conditions, clothing,
To assess whether these results are generalisable to the target pop­
ulation of shopping centre customers, we considered the interviews
conducted within each operative temperature interval as separate trial
with two possible outcomes per interview: (thermally) “comfortable”
with probability p for answers “very comfortable” to “just comfortable”
and (thermally) “uncomfortable” with probability 1-p for answers “just
not comfortable” to “very not comfortable”. The objective was to show
that at least 80% of the customers were satisfied with the thermal
environment. Using the sample proportion (i.e., the proportion of cus­
tomers in the sample who felt thermally comfortable) as estimate for p,
we thus deemed these results generalisable if the two inequalities
showed in equations (2) and (3) were met.
n̂
p (1 − ̂
p) > 5 (2)
( )
p − 0.8 2 ̂
̂ p (1 − ̂
p)
− 1
> (3)
Φ (α) n

p denotes the sample proportion, Φ− 1 the inverse of the standard

where ̂
normal cumulative distribution, and α the confidence level. The first
inequality is a normal approximation condition while the second is
equivalent to stating that the lower endpoint of the one-sided confidence
Fig. 9. PMV against TSV. The grey dashed line shows the diagonal.

Fig. 8. Bar chart of the distribution related to the thermal comfort question.

8
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

customers’ attitudes, cultural habits, etc. these temperatures.

For each day of measurement, the daily comfort temperature was
3.3. Suitability of traditional thermal comfort models for transitional calculated as shown in Equation (1):
spaces Top = 0.33 Trm + 18.8 ◦ C (1)

3.3.1. Fanger’s model where Trm is the mean running temperature calculated as per EN16798-
We assessed the applicability of Fanger’s model by comparing the 1: 2019 [31] and as introduced by Nicol and Humpreys [30].
PMVs calculated based on measured indoor conditions per each TSV As it is generally assumed, the customers were supposed to be
level for the entire dataset (Fig. 9). satisfied when their TSV was within the range of slightly cool (− 1) and
Clearly, the PMV is almost unaffected by the actual TSV. Although slightly warm (1) [20].
there is a very slight tendency for the PMV to increase with increasing Fig. 10 reports the results by day of measurement.
TSV, we refrained from drawing a regression line, because the TSV is a On April 4th, 5th and 6th the active cooling system was off in SC01.
variable of categorical nature and by rigorous standards of statistics does The mechanical ventilation system was providing just the minimum
not allow interpolation. hygienic airflow rates.
This result shows that Fanger’s model is not capable of capturing the For the measurement days in SC01 and SC02 the estimated comfort
thermal sensation reported by the customers. If Fanger’s model was temperatures are in line with the actual average operative temperatures
suitable in this context, the PMV would be a proxy for the TSV, and experienced by the customers. While for SC02 on summer period (June
binning the PMV as described in Section 2.4 and equating it with the TSV 21st- 22nd) these temperatures allow for over 90% of satisfied cus­
would lead to the relationship between TSV and PPD shown by the grey tomers, for SC01 the percentage of satisfied customers exceeds the 80%
line in Fig. 9. The narrow range of PMVs is the direct consequence of the just one day (April 6th). On April 5th, 60 customers were interviewed
narrow operation range for air conditioning in the spaces. and just 50% of them were satisfied with the thermal environment
The inadequacy of Fanger’s model in predicting thermal sensation although comfort temperature was close to average operative temper­
for transitional spaces may find explanation in the assumptions at the ature. The almost same level of operative temperature is experienced on
base of this model. The method is indeed based on the assumption that July 18th in SC03 by 52 customers and the percentage of them being
people are in a steady-state condition, which is not the case of shopping satisfied reached 100%. There are two main possible reasons concurring
centre transitional spaces. Subjects are under a constant thermal tran­ in creating this difference on customers’ thermal sensation:
sient because of moving among different zones of the shopping centre
(shops, common areas, food store, etc.). Furthermore, those interviewed • Clothing: on April 5th, customers of SC01 had an average level of
right after entering the mall experience an initial thermal sensation clothing equal to 0.78 clo, which represents a mid-season situation.
overshoot due to the temperature difference between outdoor and in­ On July 18th, the average level of clothing in SC03 was equal to 0.38
door, a parameter that is not considered by Fanger’s theory and model. clo. Therefore, experiencing the same operative temperature, the
level of satisfaction was higher when the clothing level was lower.
3.3.2. Adaptive comfort model • Outdoor-indoor temperature step: while in April the temperature
The adaptive comfort model was tested by verifying how far the increased from the outside to the inside, in July customers experi­
average operative temperatures recorded during the measurement days enced the opposite. Therefore, customers experienced different
were from the predicted comfort temperatures. The data were crossed- outdoor temperatures before entering the shopping centre.
checked with the level of satisfaction of customers with respect to

Fig. 10. Comparison between comfort temperature calculated according to the adaptive comfort model, average operative temperature, and percentage of satisfied
customers over the day of measurement. The dark green line highlights the 80% threshold of satisfied customers. (For interpretation of the references to colour in this
figure legend, the reader is referred to the Web version of this article.)

9
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

The average operative temperatures recorded in SC02, which was revealed the occurrence of a phenomenon named thermal sensation
mechanically conditioned during the measurements, were almost equal overshoot. This event consists in a variation of the thermal sensation
to the predicted comfort temperatures. The percentage of satisfied cus­ after experiencing a temperature difference while moving from outdoor
tomers was around 90%. Looking at Fig. 10 and the data from SC03, the to an indoor environment [34–36]. After moving from an environment
percentage of satisfied people was always above 80% but at lower to another that is cooler or warmer than the previous, thermal allies­
average operative temperature than the one predicted by the adaptive thesia takes effect. Thermal alliesthesia relates to the thermal pleasure
comfort model. sensation and overshoot generated by the restoration of a thermal stress
Based on this analysis, the direct application of the adaptive thermal towards stable conditions [37,38].
comfort model for transitional spaces is not recommended. This Customers moving within the shopping centre go to and from the
conclusion was reached especially for the conditions of dissatisfaction shops placed along the common areas where, most of the time, the
that the predicted comfort temperature can create in the mid-season thermal conditions are different. Therefore, customers are constantly
period. subjected to a thermal transient and they can experience two different
To derive a model to assess thermal comfort in shopping centre types of thermal sensation overshoot:
transitional spaces, it is first necessary to better understand the range of
operative temperatures judged comfortable by the customers. • A first overshoot due to outdoor-indoor temperature difference
experienced when they first enter inside the shopping centre;
4. Discussion and study limitations • Several overshoots due to the temperature difference between the
shops and the common areas.
The methodology of the field study represents both advantages and
disadvantages for research. Limitations arise from the lack of direct The thermal overshoot might generate an alliesthesial effect, which
control over the environmental variables and from the difficulty to might result in a higher percentage of satisfied customers.
precisely assess human physiological conditions. On the other hand, When a person moves from a warmer to a cooler environment, this is
field studies are of great importance to study thermal perception in a real referred to as a down-step temperature difference. In the case of shop­
environment under normal operation. ping centres, this is what happens in the summer season when customers
Physiological parameters of the customers were not directly enter the shopping centre. The link between down- or up-step temper­
measured. Therefore, the metabolic activity was assumed according to ature difference and thermal comfort perception and preference in the
the recommendation of the European standard EN ISO 7730 [27], which different seasons has been neglected in this study. Instead, we preferred
suggest a value of 1.6 met for “shopping” activity. to focus at parameters that impact the building operation like common
Besides limitations related to the field study, another limitation can areas temperatures and therefore setpoints. For what concern the down-
be found in the scale used in the questionnaire for the evaluation of the set of temperature right after entering the building, this may impact with
thermal sensation. It was not continuous, but discretized. The choice of a different extents the 33.7% of interviewees that has spent less than 10
discrete scale was mainly due to the use of a paper-based questionnaire. min indoor. Nevertheless, our intent is to present thermal sensation and
The limitation showed up when directly comparing the PMV, a contin­ comfort level of a representative sample of the customers of a shopping
uous value, with the actual thermal sensation of customers, which is an centre, which is made by users who has been inside by different amount
integer between − 3 and +3. Within the study, a direct comparison was of time, and therefore provide indication on those parameters that an
possible by categorizing the PMV. However, by doing so, a certain level energy manager can actually control to operate the systems.
of thermal sensation detail was lost.
In relation to the questionnaire, we need to consider a non- 5. Conclusions and future work
quantifiable bias due to the fact that the subjects were directly inter­
viewed by the researchers. This may also affect the results as suggested In this study we assessed the thermal perception of over 700 cus­
by McIntyre [26]. tomers through measurements and interviews performed in three Italian
The study was conducted entirely in Italian shopping centres during shopping centres. This study is one of the few works dealing with
the warm season only. This means that it reflects the thermal perception shopping centres on the Mediterranean area and compare with the work
and expectations of mostly Italian customers. The conclusions may vary of Martellotta et al. [17–19] (also dealing with Mediterranean climate),
if the study was conducted in another country. Measurements should be we analysed customers instead of workers and their perceptions over the
replicated also during the cold season to investigate the application of indoor environment. The main results of the study can be summarized as
the Fanger comfort model in transitional spaces during winter. As follows:
highlighted in Section 1, the majority of previous studies took place in
Asia or UK, and only one dealt with the Mediterranean climate [17–19] • The steady-state model of Fanger proved to be unsuitable in the
(three papers but the same case study), and even in this case the focus estimation of the thermal sensation of the customers. It tends to
was the thermal comfort of workers and not customers. In line with the overestimate the discomfort for higher operative temperature;
findings for other climates, this study shows that Fanger’s model is not • The direct application of the adaptive thermal comfort model for
suitable to control indoor transitional spaces of shopping centres in the transitional spaces is not recommended because of the discrepancy
Mediterranean area. To extend the results on all transitional spaces between predicted comfort temperature and satisfaction level in the
located in Group C climates (according to Koppen categorization), a mid-season period. The reasons for the high level of dissatisfaction
deeper analysis dealing with cultural habits and expectation is needed. were identified in the clothing level (average of 0.77 clo) combined
Another element that is worth to be discussed is the applicability of with the up-step temperature difference experienced by the
the adaptive comfort model. The peculiarities of transitional spaces may customers;
suggest treating them as a sort of hybrid areas in between a conditioned • The study showed the necessity of a tailor-made model to assess
indoor environment (shop) and the outdoor. If that were the case, the thermal comfort in transitional spaces. In order to expand the base of
adaptive comfort model could have fit the control needs of these spaces. evidence, further field studies are required, gathering together a
Instead, the results strongly pointed toward a not applicability of the conspicuous number of data covering all seasons;
model in such spaces, with customers’ answer that lacked to show that • Shopping centre customers judged operative temperatures between
link between outdoor and indoor conditions typical of studies in natu­ 23.5 and 27.5 ◦ C as comfortable. Between 24.5 and 27.5 ◦ C at least
rally ventilated buildings (Fig. 10). 88% of the respondents evaluated the indoor conditions with a TSV
Furthermore, several literature studies involving human subjects between − 1 and 1, so they are satisfied with them;

10
M. Avantaggiato et al. Building and Environment 188 (2021) 107428

• We proved statistically, at a confidence level of at least 95%, that the [8] A. Kotopouleas, M. Nikolopoulou, Thermal comfort conditions in airport terminals:
indoor or transition spaces? Build. Environ. 99 (2016) 184–199, https://doi.org/
statement “at least 80% of the customers of a shopping centre deem
10.1016/j.buildenv.2016.01.021.
an operative temperature between 24.5 and 27.5 ◦ C comfortable” [9] A.K. Mishra, R.P. Kramer, M.G.L.C. Loomans, H.L. Schellen, Development of
can be generalised to any shopping centre with comparable outdoor thermal discernment among visitors: results from a field study in the Hermitage
and indoor environmental conditions and customers’ attitudes. Amsterdam, Build. Environ. 105 (2016) 40–49, https://doi.org/10.1016/j.
buildenv.2016.05.026.
[10] V. Palma, G. Angelica, Short-term thermal history in transitional lobby spaces, PhD
Within the increasing number of papers debating on the best indoor Thesis, University of Sheffield, 2016, http://etheses.whiterose.ac.uk/12166/.
temperatures (and conditions in general) in shopping centres, this one is [11] G. Hou, An investigation of thermal comfort and the use of indoor transitional
spaces. Thesis for the Degree of Doctor of Philosophy, Cardiff University, 2016.
for sure in line with those providing evidences in favour of higher https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.723563.
temperature setpoints. Even so, we believe that there is a fundamental [12] Y. Li, S. Geng, F. Chen, C. Li, X. Zhang, X. Dong, Evaluation of thermal sensation
issue that is rarely considered in literature that concern the difference among customers: results from field investigations in underground malls during
summer in Nanjing, China, Build. Environ. 136 (2018) 28–37, https://doi.org/
between transitional spaces and the entire shopping mall (including also 10.1016/j.buildenv.2018.03.027.
shops). Although within the same building, these spaces are physically [13] J.M.Y. Tse, P. Jones, Evaluation of thermal comfort in building transitional spaces -
divided, geometrically and aesthetically distinct, and customers used field studies in Cardiff, UK, Build. Environ. 156 (2019) 191–202, https://doi.org/
10.1016/j.buildenv.2019.04.025.
them in a radically different way. This main distinction in the analysis [14] X. Du, Y. Zhang, Z. Lv, Investigations and analysis of indoor environment quality of
poses the validity of the results only for transitional spaces and highlight green and conventional shopping mall buildings based on customers’ perception,
the need of a clear distinction in future works on this topic. Build. Environ. 177 (2020), https://doi.org/10.1016/j.buildenv.2020.106851.
[15] T.F. Kwok, Y. Xu, P.T. Wong, Complying with voluntary energy conservation
Concerning future work, this study should be replicated in other
agreements (I): air conditioning in Hong Kong’s shopping malls, Resour. Conserv.
transitional spaces in different climatic and cultural contexts to extend Recycl. 117 (2017) 213–224, https://doi.org/10.1016/j.resconrec.2016.10.014.
its validity. Further, the issues mentioned in Section 4 (control over [16] P. Lu, J. Li, Acceptable Temperature Steps for Occupants Moving between Air-
environmental variables; assessment of customers’ physiological con­ Conditioned Main Space and Naturally Ventilated Transitional Space of Building,
vol. 182, Building and Environment, 2020, https://doi.org/10.1016/j.
ditions; use of a continuous thermal sensation scale in the questionnaire) buildenv.2020.107150.
should be addressed in future studies by, e.g., refining the design of [17] S. Della Crociata, F. Martellotta, A. Simone, A measurement procedure to assess
experiment and complementing the studies in the field with appropriate indoor environment quality for hypermarket workers, Build. Environ. 47 (2012)
288–299, https://doi.org/10.1016/j.buildenv.2011.07.011.
tests in controlled environments or the lab. Finally, the relationship [18] A. Simone, S. Della Crociata, F. Martellotta, The influence of clothing distribution
between down- or up-step temperature differences and thermal comfort and local discomfort on the assessment of global thermal comfort, Build. Environ.
perception as well as potential thermal alliesthesia effects experienced 59 (2013) 644–653, https://doi.org/10.1016/j.buildenv.2012.10.001.
[19] F. Martellotta, A. Cannavale, M. D’Alba, S. Della Crociata, A. Simone, Optimization
by humans in transitional spaces should be investigated. of indoor environment quality for hypermarket workers: from subjective response
to objective design criteria, Energy Procedia 101 (2016) 272–279, https://doi.org/
10.1016/j.egypro.2016.11.035.
[20] P.O. Fanger, Thermal Comfort. Analysis and Applications in Environmental
Declaration of competing interest Engineering, Danish Technical Press, Copenhagen, DK, 1970.
[21] Daniel Albuquerque, Mateus Nuno, Marta Avantaggiato, Guilherme Carrilho da
The authors declare that they have no known competing financial Graça, Full-scale measurement and validated simulation of cooling load reduction
due to nighttime natural ventilation of a large atrium, Energy Build. 224 (2020).
interests or personal relationships that could have appeared to influence
[22] R. de Dear, G. Brager, Developing an adaptive model of thermal comfort and
the work reported in this paper. preference, Build. Eng. 104 (1) (1998) 145–167.
[23] EN ISO 7726, Ergonomics of the Thermal Environment. Instruments for Measuring
Physical Quantities, 1998.
Acknowledgements [24] M.A. Humphreys, The optimum diameter for a globe thermometer for use indoors,
Ann. Occup. Hyg. 20 (1977) 135–140.
The research leading to these results has received funding from the [25] A. Simone, J. Babiak, M. Bullo, G. Landkilde, B.W. Olesen, Operative temperature
control of radiant surface heating and cooling systems. Proceedings of: Clima 2007
European Community’s Seventh Framework Programme (FP7/ Wellbeing Indoors, Finland, Helsinki, June .
2007–2013) under grant agreement no 608678. The authors thank the [26] D.A. McIntyre, Indoor Climate, Applied Science Publishers, London, 1980.
Department of Innovation, Research and University of the Autonomous [27] EN ISO 7730, Ergonomics of the Thermal Environment – Analytical Determination
and Interpretation of Thermal Comfort Using Calculation of the PMV and the PPD
Province of Bozen/Bolzano for covering the Open Access publication Indices and Local Thermal Comfort Criteria, International Standards Organization,
cost. Geneva, 2005.
[28] M. Kottek, J. Grieser, C. Beck, B. Rudolf, F. Rubel, World Map of the Köppen-Geiger
climate classification updated, Meteorol. Z. 15 (3) (2006) 259–263.
References [29] http://www.commonenergyproject.eu/uploads/deliverable/file/1/WP2_D2.1_2
0141130_P05_Shopping_malls_features_in_EU-28_and_Norway_NP.pdf.
[1] Y. Zhang, J. Zhang, H. Chen, X. Du, Q. Meng, Effects of step changes of temperature [30] F. Nicol, M. Humpreys, Derivation of the adaptive equations for thermal comfort in
and humidity on human responses of people in hot-humid area of China, Build. free- running buildings in European standards EN15251, Build. Environ. 45 (2010)
Environ. 80 (2014) 174–183, https://doi.org/10.1016/j.buildenv.2014.05.023. 11–17, https://doi.org/10.1016/j.buildenv.2008.12.013.
[2] K. Jitkhajornwanich, A. Pitts, Interpretation of thermal responses of four subject [31] EN 16798-1, Energy Performance of Buildings - Ventilation for Buildings - Part 1:
groups in transitional spaces of buildings in Bangkok, Build. Environ. 37 (2002) Indoor Environmental Input Parameters for Design and Assessment of Energy
1193–1204, https://doi.org/10.1016/S0360-1323(01)00088-9. Performance of Buildings Addressing Indoor Air Quality, Thermal Environment,
[3] C. Chun, A. Kwok, A. Tamura, Thermal Comfort in Transitional Spaces - Basic Lighting and Acoustics, 2019.
Concepts: Literature Review and Trial Measurement, Building and Environment, [32] M. Schweiker, comf. An R package for thermal comfort studies, The R Journal 8/2
2004, pp. 1187–1192, https://doi.org/10.1016/j.buildenv.2004.02.003. (December 2016).
[4] A. Pitts, J. Bin Saleh, S. Sharples, Building Transition Spaces, Comfort and Energy [33] Weather Underground, Weather History and Data Archive. https://www.
Use. PLEA 2008- 25th Conference on Passive and Low Energy Architecture, 22nd to wunderground.com/history , 16 accessed on September 2016.
24th October 2008. Dublin, Paper No 591. [34] C.P. Chen, R.L. Hwang, S.Y. Chang, Y.T. Lu, Effects of temperature steps on human
[5] R.L. Hwang, K.-H. Yang, C.-P. Chen, S.-T. Wang, Subjective responses an comfort skin physiology and thermal sensation response, Build. Environ. 46 (2011)
reception in transitional spaces for guests versus staff, Build. Environ. 43 (12) 2387–2397, https://doi.org/10.1016/j.buildenv.2011.05.021.
(2008) 2013–2021, https://doi.org/10.1016/j.buildenv.2007.12.004. [35] H. Zhang, Human Thermal Sensation and Comfort in Transient and Non-uniform
[6] Q.J. Kwong, S.H. Tang, N.M. Adam, Thermal comfort evaluation of the enclosed Thermal Environments, PhD Thesis, University of California, Berkeley, 2003.
transitional space in tropical buildings: subjective response and computational [36] R. de Dear, J. Ring, P. Fanger, Thermal sensations resulting from sudden ambient
fluid dynamic simulation, Appl. Sci. 9 (19) (2009) 3480–3490, https://doi.org/ temperature changes, Indoor Air 3 (1993) 181–192.
10.3923/jas.2009.3480.3490. [37] M. Cabanac, Physiological role of pleasure, Science 173 (1971) 1103–1107,
[7] S.C.M. Hui, J. Jiang, Assessment of thermal comfort in transitional spaces, in: https://doi.org/10.1126/science.173.4002.1103.
Proceedings of the Joint Symposium 2014: Change in Building Services for Future, [38] T. Parkinson, R. de Dear, Thermal pleasure in built environments: physiology of
25 Nov 2014, Kowloon Shangri-La Hotel, Tsim Sha Tsui East, 2014, p. 13. alliesthesia, Build. Res. Inf. 43 (2015) 288–301, https://doi.org/10.1080/
Kowloon, Hong Kong. 09613218.2015.989662.

11