lable at ScienceDirect

Building and Environment 46 (2011) 478e488

Contents lists avai

Building and Environment

journal homepage: www.elsevier .com/locate/bui ldenv

Evaluation of calculation methods of mean skin temperature for use in thermalcomfort study

Weiwei Liu a,*, Zhiwei Lian b,**, Qihong Deng a, Yuanmou Liu c

a School of Energy Science & Engineering, Central South University, Changsha, Hunan 410083, Chinab Institute of Refrigeration & Cryogenics, Shanghai Jiao Tong University, 800 Road Dongchuan, Shanghai 200240, ChinacCollege of Basic Medicine, Shanghai Jiao Tong University, Shanghai 200240, China

a r t i c l e i n f o

Article history:Received 7 July 2010Received in revised form19 August 2010Accepted 21 August 2010

Keywords:Mean skin temperatureThermal comfortHeart rate variabilitySensitivity

* Corresponding author. Tel.: þ86 731 88877175.** Corresponding author. Tel.: þ86 21 34204263; fa

E-mail addresses: wliu@mail.csu.edu.cn (W. Liu), z

0360-1323/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.buildenv.2010.08.011

a b s t r a c t

A method was established to evaluate calculation methods of mean skin temperature, in order to findappropriate ones for use in human thermal comfort study. In this method three indexes, includingreliability, sensitivity and number of measurement sites, were proposed. Under air temperatures of 21 �C,24 �C, 26 �C, and 29 �C, 22 subjects’ local skin temperatures (21 sites) and electrocardiograms weremeasured, and their thermal sensation and thermal comfort were inquired. Human heart rate variabilityindicated the physiological relation between mean skin temperature and ambient temperature for thesensitivity evaluation. Adopting the evaluation method, 26 types of mean skin temperature calculationmethods were evaluated based on the experimental data. The results indicate that a calculation methodof mean skin temperature with 10 sites is the most appropriate one, due to its high reliability, excellentsensitivity and fewer measuring sites. When it was applied to reflect thermal comfort, the performancewas good.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Mean skin temperature (MST) is an important physiologicalparameter reflecting human response to cold or thermal stimulusand states of heat exchange between human body and a thermalenvironment. Human thermal comfort is defined as being “thatcondition of mind in which satisfaction is expressed with thethermal environment” [1]. As indicated in both of its thermophy-siological definition [2] and energetic definition [3], MST playsa dominating role [4]. Therefore MST is often measured as anessential physiological parameter relating with thermal comfort(e.g. [5e8]).

Values for MSTare obtained by summing the products of a finitenumber of local skin temperatures and the correspondingweighting factors. Up to now, many MST calculation methods havebeen established from the field of physiology, distinguished by thenumber of skin temperature sites and weighting factors. In thestudies on thermal comfort, subjects’ MST was measured withone of these MST calculation methods. For example, Bulcao useda 10-site weighed MST calculation method [5], Gagge adopted anaverage of 10 sites as mean skin temperature [6], and Hasebe and

x: þ86 21 34206814.wlian@sjtu.edu.cn (Z. Lian).

All rights reserved.

Huizenga chose a 7-site method [7,8]. The MSTcalculation methodsthemselves cause differences in the MST values. That is to say ifdifferent one was used, the results might be distinct. However, inthese studies no reason was given to explain why the MST calcu-lation method was chose. Considering the importance of anappropriate MST calculation method in obtaining a reasonableresult, it is necessary to compare various MST calculation methodsand find out which are suitable for use in thermal comfort study.

In the present work, a method was established to evaluatedifferent MST calculation methods considering the effect ofambient temperature on skin temperature, which is the mostimportant environmental factor affecting human thermal comfort.And also, the most appropriate one for future use in study onthermal comfort was discussed.

2. Methods

2.1. Subjects

12 male and 10 female college students (mean � SEM of age:23.9� 0.4 years, height: 170.6 � 1.1 cm, weight: 61.2 � 1.6 kg) wererecruited for the experiment. All subjects were healthy non-smokers who were not taking prescription medication and had nohistory of cardiovascular disease. All protocols were approved bythe university’s ethics committee and conformed to the guidelines

W. Liu et al. / Building and Environment 46 (2011) 478e488 479

contained within the Declaration of Helsinki. Verbal and writteninformed consent was obtained from each subject prior to theparticipation in the protocol. Subjects were asked to avoid caffeine,alcohol, and intense physical activity at least 12 h prior to eachexperimental session.

2.2. Instrumentation

Subjects’ local skin temperatures were measured with copper-constantan thermocouples attached to the skin measuring sites, asshown in Fig. 1. During the measurement, the thermocouples werelinked to a multi-channel data collector with internal referencejunction (KEITHLEY 2700, Keithley Instruments, USA), and skintemperatures were automatically recorded to a computer via thedata collector, at 5 s interval. Before the measurement, all thethermocouples were calibrated against a standard mercury ther-mometer with precision of 0.1 �C. Error of the thermocouples was0.2 �C.

Subjects’ electrocardiogram (ECG) was recorded by a Powerlab8/30 system (AD Instruments, Australia). For ECGmeasurement, thestandard bipolar limb leads (ML-1340, AD Instruments, Australia)and the adhesive ECG pads (MLA-1010, AD Instruments, Australia)were linked to the data acquisition system (Powerlab 8/30, ADInstruments, Australia) through a bioamplifier (ML-132, AD Instru-ments, Australia). The Powerlab 8/30 system was calibrated before

Fig. 1. Measuring sites of skin temperature. A forehead, B left check, C left neck, D rightupper arm, E left elbow, F left forearm, G left palm, H right hand, I left hand, J left back,K left chest, L left lumbar, M left abdomen, N left buttocks, O anterior thigh, P posteriorthigh, Q anterior calf, R posterior calf, S left foot, T right foot, U left sole.

the experiment. The frequency of sampling for ECG was set at 400times per s.

The ambient temperature was measured with a standardmercury thermometer (Shanghai Huo er Co, China). Indoor airvelocity was tested using an anemoscope (TSI Compuflow 8585,E&E Process Instrumentation, Canada). And the relative humidity ofindoor air was measured with a dryewet bulb thermometer(Shanghai Huachen Medical Instruments Co, China). The meanradiant temperature was obtained according to Eq. (1), where theblack-bulb temperature was measured by a standard thermometer(D 150 mm, Shanghai Huo er Co, China).

Tr ¼h�tg þ 273

�4þ0:4� 108�tg � ta

�5=4i1=4�273 (1)

where Tr is mean radiant temperature, tg black-bulb temperatureand ta is air temperature. The indoor environmental parametermeasurement site was located in the center of the plane at 0.6 mheight (near the subject).

2.3. Experimental protocol

In order to consider the effect of the order of air temperaturechange on MST, the subjects were divided into two groups and ineach group a different combination of ambient temperatures wasdesigned. Group 1 (6 men and 5 women, G1) was exposed to anindoor environment with the temperature order of 21 �C, 24 �C,26 �C and 29 �C, for group 2 (6 men and 5 women, G2) the orderwas 29 �C, 26 �C, 24 �C and 21 �C. The experiment (all the fourindoor temperatures) was done for only one subject on a single day.During the experiment the subject was blind to the exposuretemperature.

The summer clothing (vest and shorts) with a total clo of about0.3 [1] was compulsory for every subject. Under each ambienttemperature, a subject’s local skin temperatures at 21 measure-ment sites and ECG were recorded for 5 min. After the measure-ment, he or she was asked to complete a questionnaire aboutthermal sensation and comfort. During the exposure andmeasurement the subjects were asked to lie quietly in a bed butkeep awake.

After the measurement at one exposure temperature wasfinished, the exposure temperature was set to the next value. Theexperimental temperature reached the new value in 20 min. Afterthe regulation of the environmental temperature began, the subjectaccommodated to the new exposure temperature. The experi-mental data in earlier studies indicated that mean skin temperatureand thermal sensation became stable within 40 min under a newthermal environment (the change of environmental temperaturewas less than 10 �C) [8e10]. In this study, the measurements weremade after the subject had stayed at a steady ambient temperaturefor at least 40 min.

The experiment was performed in a climate chamber (see Fig. 2)duringMay. All measurements were carried out between 13:00 (1 hafter lunch) and 17:30. As shown in Fig. 2, there was only a windowin the climate chamber and no direct solar radiation entered. Theair temperature was controlled using a wall-mounted air condi-tioner. The indoor air velocity near the subjects was kept under0.05 m/s, and the relative humidity of air was not dependentlycontrolled.

2.4. Thermal sensation and thermal comfort

At different air temperatures, an occupant may have distinctthermal sensation. Here, the ASHARE 7-point scale was used toassess the subjects’ thermal sensation (Fig. 3). The ASHARE scale is

Fig. 2. Schematic diagram of the climate chamber.

W. Liu et al. / Building and Environment 46 (2011) 478e488480

based on ameasure of howwarm or cool the occupant feels. Duringthe experiment, a thermal sensation scale was obtained by ques-tionnaire, inwhich the subject was required to choose a point (from�3 to þ3) according to the thermal sensation.

Feelings of thermal comfort may also have to do with expecta-tion, adaptation to conditions besides other factors [1]. Therefore,information about thermal sensation alone is not sufficient toreflect the thermal comfort. Subjects’ thermal comfort level(comfortable or uncomfortable) at a particular thermal sensationwas inquired, and also, the subjects were asked to describewhetherthey sweat when they felt warm or hot.

2.5. Human heart rate variability (HRV)

HRV is measured by determining the constantly changingtemporal distance between succeeding ReR intervals. It is essen-tially based on the antagonistic oscillatory influences of thesympathetic and parasympathetic nervous system on the nodussinuatrialis of the heart [11]. HRV is an integrating measurement

Fig. 3. Seven step scales of human thermal sensation.

variable that reflects, especially under resting conditions, the pre-vailing balance of vagus and sympathetic tone [12].

There are three main spectral frequency bands in the frequencydomain analysis of HRV. The power spectrum of the high frequencyband (HF: 0.15 e 0.40 Hz) is affected by activity of the vagalnerve, while the power spectrum of the low frequency band (LF:0.04 e 0.15 Hz) mainly relates to activity of the sympathetic nerve[13e16]. Therefore, in analysis of HRV, the ratio of absolute power inLF and HF bands (LF/HF ratio) is considered an indicator of sym-pathetic:parasympathetic balance.

Human thermoregulation is closely related to thermal comfort.When a person feels thermally uncomfortable, the thermoregula-tion, including vasoconstriction and sweating, is strong. While hefeels thermally comfortable, the thermoregulation is minimized [1].Human thermoregulation is controlled by the autonomic nervoussystem [17]. Activation of the sympathetic nerve leads to vasocon-striction or exudation of sweat gland. In the analysis of HRV, theLF/HF ratio indicates the activity of the sympathetic nerve. There-fore, the LF/HF ratio might be used to reflect thermal comfort [7,18].The LF/HF ratio was calculated based on a 5-min record of ECG.

2.6. Mean skin temperature calculation method

Human MST (Tsk) can be estimated by the following generalformula:

Tsk ¼Xn

i¼1

kiti (2)

where ti is local skin temperature, ki is the corresponding weight-ing factor and n is the number of skin measuring sites. Usually, theweighting factor ki is the fraction of total body area with temper-ature ti.

The MST calculation methods evaluated in the present work arelisted in Table 1, with the measuring sites and the correspondingweighting factors (see Refs. [19e24]). Here, these calculationmethods were named according to the number of the measure-ment sites. For example, four methods with 10 sites were called10a, 10b, 10c and 10d, respectively.

These MST calculation methods can be classified into two types:(1) weighted methods using constant weighting factors accordingto relative regional surface area of specific measuring sites; (2)unweighted methods using the same weighting factors for all themeasurement sites, result of which is the average of the whole localskin temperatures. The weighted calculation methods containmore information about human body surface temperature distri-bution than the unweighted methods [19], which is useful forstudies of thermal comfort. Therefore, the emphasis is laid on theweighted calculationmethods. Only two unweightedmethods (10dand 14) were considered.

2.7. Evaluation indexes for MST calculation methods

Air temperature is the most important environmental factoraffecting human thermal comfort. When relatingMSTwith thermalcomfort, not only the value of MST under one air temperature butalso the difference in MST between different air temperatures needto be seriously investigated. Considering this specific requirementof thermal comfort study, three indexes are proposed to evaluatethe MST calculation methods.

(1) Reliability. It reflects the accuracy of MST obtained by a calcu-lation method, compared with a reference value under thesame thermal condition.

Table 1Mean skin temperature calculation methods.

Method U T S R Q P O N M L K J I H G F E D C B A

3a 0.36 0.5 0.143b 0.25 0.5 0.254a 0.18 0.33 0.34 0.154b 0.2 0.2 0.3 0.34c 0.28 0.28 0.16 0.285a 0.20 0.18 0.50 0.05 0.075b 0.39 0.175 0.175 0.19 0.076a 0.32 0.19 0.19 0.05 0.11 0.146b 0.186 0.186 0.186 0.186 0.107 0.1496c 0.32 0.19 0.19 0.05 0.11 0.147a 0.07 0.13 0.19 0.35 0.05 0.14 0.077b 0.206 0.172 0.166 0.162 0.114 0.082 0.0988a 0.16 0.23 0.11 0.11 0.11 0.11 0.085 0.0858b 0.2 0.19 0.175 0.175 0.05 0.07 0.07 0.078c 0.08 0.15 0.17 0.1 0.11 0.06 0.12 0.218d 0.15 0.12 0.12 0.08 0.09 0.12 0.13 0.199 0.06 0.13 0.19 0.18 0.18 0.05 0.07 0.07 0.0710a 0.05 0.15 0.125 0.125 0.125 0.125 0.06 0.07 0.07 0.110b 0.06 0.115 0.19 0.095 0.095 0.19 0.045 0.06 0.09 0.0610c 0.07 0.13 0.19 0.12 0.12 0.12 0.05 0.06 0.08 0.0610d 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.111 0.07 0.13 0.19 0.09 0.09 0.09 0.09 0.05 0.07 0.07 0.0612 0.07 0.065 0.065 0.095 0.095 0.0875 0.0875 0.0875 0.0875 0.05 0.14 0.0714 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.07115 0.0325 0.0325 0.0625 0.0625 0.1025 0.1025 0.2 0.18 0.0225 0.0225 0.025/0.025 0.035/0.035 0.0617 0.0305 0.0305 0.0875 0.0875 0.0875 0.0875 0.063 0.063 0.063 0.063 0.063 0.025 0.025 0.075 0.075 0.037 0.037

Method Proposer Year Reference

3a Burton 1934 [19,20]3b e e [21]4a Newburgh & Spealman 1943 [20]4b Ramanathan 1964 [19,20]4c ISO 1992 [22]5a e e [23]5b Houdas 1982 [20]6a e e [23]6b Teichner 1958 [19]6c Palmes & Park 1947 [19,20]7a Hardy & Dubios 1938 [19,20]7b e e [23]8a e e [23]8b Gagge & Nishi 1977 [20]8c Nadel 1973 [20]8d Crawshaw 1975 [20]9 e e [23]10a QREC 1943 [19,20]10b e e [23]10c Colin & Houdas 1982 [20]10d Stolwijk & Hardy 1966 [20]11 e e [23]12 Hardy & Dubios 1938 [20]14 ISO 1992 [24]15 e e [23]17 e e [23]

Values are the weighting factors of the mean skin temperature calculation methods. A forehead, B left check, C left neck, D right upper arm, E left elbow, F left forearm, G left palm, H right hand, I left hand, J left back, K left chest,L left lumbar, M left abdomen, N left buttocks, O anterior thigh, P posterior thigh, Q anterior calf, R posterior calf, S left foot, T right foot, U left sole. e means the information about Proposer and Year was uncertain.

W.Liu

etal./

Buildingand

Environment

46(2011)

478e488

481

W. Liu et al. / Building and Environment 46 (2011) 478e488482

(2) Sensitivity. It means whether an MST calculation method cansensitively reflect the change of skin temperature with airtemperature.

(3) Number of measurement sites. There are two reasons to list itas an evaluation index. One is that too many measuring sitesmight affect subjects’ thermal feelings. The other is thatnumerous measuring sites are likely to reduce the efficiency ofskin temperature measurement in field study of thermalcomfort. Moreover, restricted by instruments and experimentalconditions, a large number of measurement sites are impos-sible. Thus, under the precondition of high reliability andsensitivity, the number of measurement sites is expected to beas small as possible.

When evaluating an MST calculation method, the three indexesare considered together. Reliability relates to the accuracy of anMSTcalculationmethod, sensitivity connects with thermal comfort,and number of measurement sites aims at easy measurement.A suitable MST calculation method should have high reliability,excellent sensitivity and fewer measuring sites.

2.8. Method of reliability evaluation

First, a reference value of human MST is required. And then, thevalues of MST calculated by different calculation methods arecompared with the reference value. Based on the comparison,reliability of the MST calculation methods can be obtained.

Both consistency (test of significance) and agreement frequencyare the existing methods to compare accuracy of various MSTcalculation methods. Population consistency is estimated by test ofsignificance, which can reflect statistical difference between twopopulations. Agreement frequency indicates agreement degreebetween two samples [19]. Here, the agreement frequency (AF) wasdefined as,

Table 2Experimental conditions.

Case Air temperature(�C)

Relativehumidity (%)

Mean radianttemperature (�C)

1 21.2 � 0.1 42.6 � 1.2 21.8 � 0.12 24.0 � 0.0 50.6 � 1.3 24.7 � 0.13 26.0 � 0.1 56.3 � 1.2 26.5 � 0.14 29.0 � 0.0 56.9 � 1.0 29.5 � 0.1

Values are means � SEM.

AF ¼ number of individuals agreeing with the reference valuetotal number of individuals in a sample

� 100% (3)

where a sample means a group of MST values obtained using onecalculation method.

The target of the present work is to find the most appropriatecalculation method for thermal comfort study. Therefore, consis-tency and agreement frequency are used together, in order tostrictly choose the method with high reliability. Both high consis-tency and agreement frequency are required for a high reliabilitycalculation method.

2.9. Method of sensitivity evaluation

Human MST varies with ambient temperature, which greatlyimpact human thermal comfort. However, it’s not easy to know thetrue significance of variation inMSTwhen air temperature changes,before the sensitivity of MST calculation methods were evaluated.The change of skin temperature relates closely to variation of skinblood flow rate. Usually, larger skin blood flow rate leads to higherskin temperature. Thus, according to the change of skin blood flowrate, the significance of variation in MST can be estimated.

Variation of skin blood flow rate is the result of vasomotion(vasoconstriction or vasodilation), which is an important physio-logical process of human thermoregulation [1]. As mentionedbefore, HRV can reflect the relation between air temperature,thermal comfort and thermoregulation. Therefore the significanceof difference in skin temperatures at different air temperature can

be reflected by the results of thermoregulation based on the anal-ysis of HRV. When performing the sensitivity evaluation, it isinspected whether the variation in MST obtained by a calculationmethod accords with the physiological mechanism revealed by theanalysis of HRV.

2.10. Reference value of MST for reliability evaluation

The reference value of MST is required to be most accurate toreflect the true value of MST, based on which the reliability ofvariousMSTcalculationmethods can be evaluated. According to thedefinition of MST, it is reasonable to consider that accuracy of MSTis increased in proportion to the number of skin temperaturemeasurement sites. In Mitchell andWyndham’s study [19], theMSTcalculated by a 15 sites (the maximum in their study) method wastaken as the reference value. And, Choi et al. adopted the mean of10,841 skin temperatures measured by infrared thermography asthe reference value [20].

In this study, the MSTcalculated by amethod with 17measuringsites (the calculation method 17 in Table 1) was taken as the refer-encevalue, considering itsmaximumsites and rational arrangementof measuring sites on human skin. And also, calculated with themethod 17, the variation in MST with thermal comfort level agreedwith the physiological mechanism revealed by the analysis of HRV(as reflected by the results of sensitivity evaluation).

2.11. Statistical analysis

Experimental datawerepresentedasmeans�SEM.Prior toa testof significance, Levene’s test of homogeneity of variance was done.For equal variances, the samples were compared using Dunnett’s ttwo-sided or StudentNewmaneKeuls tests (a two-wayANOVA). Forunequal variances, Dunnett’s C test (a two-way ANOVA) was used.Among these methods, Dunnett’s t two-sided test was used tocompare multiple groups of samples with a reference group, and

Student NewmaneKeuls test or Dunnett’s C test was performed formultiple range test. The level of significance was set at P < 0.05.

3. Results

3.1. Environmental parameters

As listed in Table 2, under every experimental condition, theenvironmental parameters were kept steady. The air was calmwithits velocity less than 0.05 m/s.

3.2. Thermal sensation and thermal comfort level

Table 3 shows data for thermal sensation and comfort. For thesubjects in groups 1 and 2, it seems that an indoor air temperature of

Table 3Data for thermal sensation and thermal comfort.

Experimentalcondition

Hot (þ3) Warm (þ2) Slightlywarm (þ1)

Neutral (0) Slightlycool (�1)

Cool (�2) Cold (�3)

C U C U C U C U C U C U C U

Group 1 21 �C 0 0 0 0 0 0 0 0 0 0 2 4 0 524 �C 0 0 0 0 0 0 2 0 4 0 5 0 0 026 �C 0 0 1 0 2 0 8 0 0 0 0 0 0 029 �C 0 0 4 7 0 0 0 0 0 0 0 0 0 0

Group 2 29 �C 0 0 3 8 0 0 0 0 0 0 0 0 0 026 �C 0 0 0 0 0 0 10 0 1 0 0 0 0 024 �C 0 0 0 0 0 0 0 0 1 0 10 0 0 021 �C 0 0 0 0 0 0 0 0 0 0 0 5 0 6

Total 0 0 8 15 2 0 20 0 6 0 17 9 0 11

Values are the number of subjects with corresponding thermal sensation and comfort. C means comfortable and U uncomfortable. For group 1 the environmental temperaturechanged from 21 to 29 �C and for group 2 from 29 to 21 �C.

W. Liu et al. / Building and Environment 46 (2011) 478e488 483

21 �Cmademost (9 in group 1 and 11 in group 2) feel uncomfortablewith the sensationof cool or cold, and29 �Cwas a temperature likelyto provoke a sensation of warm, with 7 (group 1) and 8 (group 2)subjects feeling uncomfortable, respectively. At temperatures of 24and 26 �C, all subjects felt thermal comfortable.

According to the subjects’ reflection, when they were uncom-fortable warm at an air temperature of 29 �C, sweat occurredslightly at the locations of hands and forehead, as well as on thechest or back.

3.3. LF/HF ratios at environmental temperatures

Subjects’ LF/HF ratios at the four environmental temperatureswere obtained, as depicted in Fig. 4.

Results of Student NewmaneKeuls test indicated that the LF/HFratios at 21 and 29 �C were higher than those under 24 and 26 �C(n ¼ 11, P < 0.05). However, the difference between the two LF/HFratios at24 and26 �Cwasnot statistically significant (n¼11, P>0.05).

3.4. Gender difference in thermal sensation and MST

Table 4 compares male and female subjects’ thermal sensationand MST under each experimental temperature. The data of MSTwere the reference values calculated using the method 17. Asshown in Table 4, female mean skin temperature was smaller thanthat of male at each experimental temperature. Correspondingly,females’ mean thermal sensation was lower than males’ (exceptunder the temperature of 26 �C). However, the differences inthermal sensation and MST between females and males were notstatistically significant (P > 0.05) almost at all the experimentalconditions, except the temperature of 29 �C in group 1. Thereforethe data of males and females were combined for further analysis.

3.5. Evaluation of reliability

3.5.1. Agreement frequencyConsidering the precision of the thermocouples (�0.2 �C),

�0.4 �C was taken as the limit of agreement. According to equation(3), the agreement frequencies of the MST calculation methodswere obtained, as shown in Fig. 5.

It can be seen from Fig. 5 that the agreement frequencies ofthe MST calculation methods with fewer than 9 measuring siteswere not higher than 60% (except for the method 6c). For themethods 8c e 14, there was a clear trend that the agreementfrequency rose with the increasing number of skin measuringsites. For the method 15, its agreement frequency decreasedslightly. Both the methods 12 and 14 had the highest agreementfrequency of 95%.

3.5.2. Population consistencySignificance of the difference between the values of MST

obtained by a calculation method and the reference values by thecalculation method 17 were tested using Dunnett’s t two-sidedmethod under each air temperature.

According to the results shown in Tables 5 and 6, the meandifference of up to 0.8 �C indicates that the MST obtained by twocalculation methods are insignificantly different statistically(P> 0.05). However, for the application in study of thermal comfort,the difference of 0.8 �C may be too large. Therefore, the meandifference between oneMSTcalculationmethod and themethod 17is also required within the agreement limit of �0.4 �C (consideringthe precision of the thermocouples). Under this condition, thecorresponding value of P is larger than 0.8.

It can be known that the MST calculation methods 4b, 10c, 12,and 14 had high population consistency, which possess the meandifference of less than 0.4 �C at the entire four air temperatures.

3.5.3. Reliability of the MST calculation methodsIn most cases, higher population consistency means higher

agreement frequency, such as the calculation methods 12 and 14.However, there is a disagreement between the two indexes forsome MST calculation methods, such as the methods 4b, 11 and 15.Thus, evaluating the reliability of the MSTcalculation methods onlyby one index is insufficient. It is more reasonable to consider thepopulation consistency and the agreement frequency together.

As to the MST calculation methods 10c, 12, and 14 with highpopulation consistency, they also have a high value of agreementfrequency (>0.8). Therefore, besides the reference calculationmethod 17, they are regarded as having a high accuracy.

In Mitchell and Wyndham’s research [19], they found that thecalculation method 4b using only four measurement sites ratedsurprisingly well (with agreement frequency of 64%). In the presentwork, the similar results were obtained for themethod 4bwith highpopulation consistency and an agreement frequency of 60%. There-fore, the calculation method 4b was regarded as a reliable one, too.

3.6. Evaluation of sensitivity

3.6.1. HRV, MST and thermal comfortRelatedwith the subjects’ thermal comfort level, the variation in

LF/HF ratio (see Fig. 4) indicated the following physiologicalmechanism.

Under the lower environmental temperature (21 �C), the subjectsfelt uncomfortable cool or cold. Theactivityof the sympatheticnervewas enhanced (the LF/HFratiowas significantlyhigher), as a result ofwhich the skin blood flowwas reduced (vasoconstriction in humanskin) to conserve body heat [1]. Thus, the skin temperature

Fig. 4. Low versus high frequency ratio (mean � SEM) for heart rate variability.Evaluated by the ECG over 5 min in two groups of subject at different indoortemperatures. For each group, n ¼ 11. For group 1, the temperature changed from 21 to29 �C; and for group 2 from 29 to 21 �C. LF/HF ratio means the ratio of absolute powerin low frequency and high frequency bands.

Table 4Thermal sensation and mean skin temperature for males and females.

Experimentalconditions

Mean TS Mean MST (�C)

Male Female Male Female

Group 1 21 �C �2.0 �2.60 32.19 31.5824 �C �0.67 �1.20 32.79 32.4026 �C 0.17 0.40 33.48 32.7529 �C 2.0 1.20* 33.94 33.18*

Group 2 29 �C 1.83 1.60 34.54 34.1126 �C �0.17 0 33.57 33.2824 �C �1.0 �1.20 32.80 32.3921 �C �2.33 �2.80 32.14 31.76

TS means thermal sensation and MST means skin temperature. * indicates signifi-cantly different (P < 0.05 in independent samples T-test) at the same experimentaltemperature. For group 1 the environmental temperature changed from 21 to 29 �Cand for group 2 from 29 to 21 �C.

Fig. 5. Agreement frequencies of the MST calculation methods.

W. Liu et al. / Building and Environment 46 (2011) 478e488484

significantly decreased. In the higher environmental temperature(29 �C), most subjects felt uncomfortable warm. The significantlyhigher value of LF/HF ratio (compared with the values in 26 �C)implied that the sympathetic nerve was activated, which inducedthat the sweat gland began to pump perspiration onto the skinsurface to cool the skin to increase heat loss from the core [1]. Underthis condition, vasodilation of skin blood vessels induced thesignificantly increasing of skin temperature.

Therefore, it is reasonable to conclude that the subjects’ MST inthe uncomfortable cool/cold (21 �C) or warm environment (29 �C)should be significantly different from the values in the comfortablethermal environment (24 and 26 �C).

More detailed discussion can be found in our published work onthe relationship between HRV and thermal comfort [18].

3.6.2. Sensitivity of the MST calculation methodsHere, the sensitivity evaluation was carried out only on the four

MST calculation methods, 4b, 10c, 12 and 14, which have highreliability, as well as the reference method 17. Significance ofdifference in the values of MST under the whole indoor airtemperatures was tested using Dunnett’s C method for the method4b (unequal variances) and Student NewmaneKeuls method forthe other methods.

For indoor air temperature decreasing, the results in Table 7indicated that the difference in MST under the entire indoor airtemperatures was significant, when using those five methods.However, for air temperature increasing, the significant differencein MST between the comfortable air temperature (26 �C) and theuncomfortable air temperature (29 �C)was obtained (see Table 8), ifthe MST calculation methods 10c and 17 were used. Therefore, onlythe MST calculation methods 10c and 17 have enough high sensi-tivity to accurately reflect the variation in human MST with the airtemperature.

The high sensitivity of the method 17 also indicated that it wasreasonable to use this method to get the reference value of MST forreliability evaluation.

4. Discussion

The MST calculation methods have been evaluated in the pastonly on their reliability [e.g. 19,20]. This quality was also reflectedby the index of reliability in the present study. Five MST calcula-tion methods including 4b, 10c, 12, 14 and the reference method 17

Table 5Significance of difference in the mean skin temperature calculation methods for temperature increasing.

Methods (I) Reference (J) Indoor air temperature (�C)

21 24 26 29

3a 17 0.179 (0.71) 0.011 (0.65)* 0.082 (0.52) 0.695 (0.24)3b 17 0.000 (�1.32)* 0.000 (�1.200)* 0.000 (�0.842)* 0.004 (�0.543)*4a 17 0.984 (0.33) 0.065 (0.54) 0.609 (0.33) 1.000 (0.04)4b 17 1.000 (0.18) 0.760 (0.30) 1.000 (0.12) 1.000 (0.04)4c 17 0.057 (0.84) 0.003 (0.73)* 0.009 (0.65)* 0.009 (0.51)5a 17 0.003 (1.10)* 0.000 (1.07)* 0.000 (0.87)* 0.001 (0.60)*5b 17 0.925 (0.39) 0.243 (0.44) 0.725 (0.30) 0.999 (0.14)6a 17 0.056 (0.84) 0.000 (0.83)* 0.000 (0.90)* 0.000 (0.66)*6b 17 0.315 (0.63) 0.018 (0.62)* 0.104 (0.50) 0.683 (0.25)6c 17 0.670 (0.49) 0.049 (0.56)* 0.330 (0.40) 0.909 (0.20)7a 17 0.999 (0.26) 0.800 (0.29) 0.987 (0.21) 1.000 (0.10)7b 17 0.532 (0.54) 0.008 (0.67)* 0.161 (0.46) 0.698 (0.24)8a 17 0.425 (0.58) 0.030 (0.59)* 0.234 (0.433) 0.739 (0.235)8b 17 0.146 (0.736) 0.028 (0.598)* 0.210 (0.44) 0.331 (0.31)8c 17 0.003 (1.11)* 0.000 (0.87)* 0.003 (0.71)* 0.000 (0.64)*8d 17 0.418 (0.59) 0.330 (0.41) 0.709 (0.31) 0.371 (0.30)9 17 0.989 (0.32) 0.258 (0.43) 0.253 (0.43) 0.192 (0.35)10a 17 1.000 (0.19) 0.988 (0.21) 1.000 (0.14) 1.000 (0.12)10b 17 1.000 (0.19) 0.277 (0.43) 1.000 (0.14) 0.848 (0.21)10c 17 0.991 (0.31) 1.000 (0.07) 1.000 (�0.02) 1.000 (0.06)10d 17 1.000 (0.22) 1.000 (0.11) 1.000 (0.10) 1.000 (0.13)11 17 1.000 (0.25) 0.974 (0.23) 0.980 (0.22) 0.973 (0.17)12 17 1.000 (0.03) 1.000 (0.05) 1.000 (0.14) 1.000 (0.10)14 17 1.000 (0.14) 1.000 (0.12) 1.000 (0.10) 1.000 (0.04)15 17 0.988 (0.32) 0.489 (0.37) 0.627 (0.33) 0.449 (0.29)

Values are the probability (P) of significance test (n ¼ 11). Numbers in parentheses are the mean difference (IeJ, �C). * indicates significantly different (P < 0.05 in Dunnett’s ttwo-sided test) from the reference value of MST obtained by the method 17. The experimental air temperature changed from 21 to 29 �C.

W. Liu et al. / Building and Environment 46 (2011) 478e488 485

were evaluated as high reliability based on the populationconsistency and agreement frequency, which agreed with Mitchell& Wyndham’s research [19] and Choi et al.’s using agreementfrequency [20]. However, the criterion in this study was morerigorous, because not only the total agreement frequency but alsothe population consistency (by setting a limit of �0.4 �C for themean difference) at different air temperatures was investigated.According to the results, it seems that the reliability of an MST

Table 6Significance of difference in the mean skin temperature calculation methods for temper

Methods (I) Reference (J) Indoor air temperature (�C

21

3a 17 0.997 (0.22)3b 17 0.000 (�1.0)*4a 17 1.000 (0.16)4b 17 1.000 (�0.05)4c 17 0.002 (0.87)*5a 17 0.023 (0.71)*5b 17 0.451 (0.44)6a 17 0.000 (0.96)*6b 17 0.502 (0.42)6c 17 0.969 (0.27)7a 17 1.000 (�0.08)7b 17 0.472 (0.43)8a 17 0.147 (0.56)8b 17 1.000 (�0.06)8c 17 0.030 (0.69)*8d 17 1.000 (0.19)9 17 0.053 (0.65)10a 17 1.000 (�0.06)10b 17 0.690 (0.37)10c 17 1.000 (�0.11)10d 17 1.000 (0.14)11 17 0.217 (0.52)12 17 0.989 (0.24)14 17 1.000 (0.19)15 17 0.585 (0.40)

Values are the probability (P) of significance test (n ¼ 11). Numbers in parentheses are thtwo-sided test) from the reference value of MST obtained by the method 17. The experi

calculation method mainly depends on the number of measure-ment sites.

Reliability is necessary but not sufficient to warrant furtherapplication of an MST calculation method in thermal comfort.Usually, the variation in skin temperaturewith ambient air needs tobeanalyzed in studieson thermal comfort [e.g. 5,7,8,25]. The indexofsensitivityweproposed is one that has ameaning in termsof humanthermal response to his thermal environment (thermoregulation),

ature decreasing.

)

24 26 29

0.002 (0.68)* 0.996 (0.17) 0.038 (0.37)*0.000 (�0.90)* 1.000 (�0.07) 1.000 (0.06)0.218 (0.41) 1.000 (0.04) 1.000 (0.10)0.991 (0.19) 0.971 (�0.21) 1.000 (�0.07)0.000 (0.92)* 0.009 (0.60)* 0.022 (0.39)*0.000 (0.85)* 0.820 (0.26) 0.139 (0.31)0.304 (0.38) 1.000 (�0.05) 1.000 (�0.07)0.003 (0.66)* 0.142 (0.44) 0.053 (0.35)0.999 (0.16) 0.785 (�0.27) 1.000 (�0.10)1.000 (0.07) 0.700 (�0.29) 1.000 (0.00)0.361 (0.37) 1.000 (0.00) 0.409 (0.25)0.011 (0.60)* 0.683 (0.29) 0.792 (0.19)0.003 (0.67)* 0.895 (0.24) 0.933 (0.16)0.007 (0.62)* 0.887 (0.24) 0.956 (0.15)0.001 (0.72)* 0.794 (0.26) 0.427 (0.24)0.318 (0.38) 1.000 (0.04) 0.998 (0.12)0.329 (0.38) 0.462 (0.34) 0.695 (0.20)0.713 (0.29) 1.000 (0.09) 0.145 (0.31)0.120 (0.46) 0.897 (0.24) 0.458 (0.24)1.000 (0.15) 1.000 (�0.11) 1.000 (0.05)0.522 (0.33) 1.000 (0.07) 0.975 (0.14)0.760 (0.28) 0.982 (0.20) 0.906 (0.17)1.000 (0.09) 1.000 (0.08) 1.000 (0.07)0.996 (0.18) 1.000 (0.06) 1.000 (0.05)0.928 (0.23) 0.999 (0.16) 1.000 (0.09)

e mean difference (IeJ, �C). * indicates significantly different (P < 0.05 in Dunnett’s tmental temperature changed from 29 to 21 �C.

Table 7Results of Significance test for mean skin temperatures in the process of air temperature decreasing.

Method Indoor air temperature (�C) Significance group of mean skin temperature (�C)

1 2 3 4

4b 21 31.91 � 0.2424 32.80 � 0.2626 33.23 � 0.2529 34.27 � 0.14

10c 21 31.86 � 0.2524 32.76 � 0.1826 33.33 � 0.1629 34.39 � 0.15

12 21 32.21 � 0.2024 32.71 � 0.1326 33.52 � 0.1729 34.41 � 0.16

14 21 32.15 � 0.2024 32.79 � 0.1426 33.50 � 0.1729 34.40 � 0.16

17 21 31.97 � 0.2124 32.62 � 0.1726 33.44 � 0.1829 34.34 � 0.15

Values are means � SEM. of mean skin temperature. For an MST calculation method, the mean skin temperatures in the same group were insignificantly different (n ¼ 11,P > 0.05). While the mean skin temperatures in different groups were significantly different (n ¼ 11, P < 0.05).

W. Liu et al. / Building and Environment 46 (2011) 478e488486

which is an important factor relating thermal comfort [1]. Sensitivityestimateswhether anMSTcalculationmethod can accurately reflectthe variationofMSTwith air temperature.Here, theHRVwasused toindicate the physiological relation between human MST andambient temperature for the evaluation of sensitivity.

The sensitivity of an MST calculation method depends on thelocations of measurement sites. Fig. 6 illustrates the local skintemperature (means) at each measurement site under the fourexperimental temperatures. It can be observed that the change ofskin temperatures on the extremities (especially sole, foot, arm andhand) was larger than the trunk and forehead, when air tempera-ture changed. This is in accordance with the trend presented inHuizenga and Zhang’s experiment [8,25]. One possible reason isthat the skin temperatures of the trunk and forehead are directlyinfluenced by the internal temperature of body and head. Addi-tionally, the clothing (Vest and shorts in this experiment) reduces

Table 8Results of Significance test for mean skin temperatures in the process of air temperature

Method Indoor air temperature (�C) Significance group

1

4b 21 32.09 � 0.4524 32.91 � 0.2126 33.27 � 0.2129

10c 21 32.23 � 0.27242629

12 21 31.95 � 0.37242629

14 21 32.05 � 0.32242629

17 21 31.91 � 0.36242629

Values are means � SEM. of mean skin temperature. For an MST calculation method, thP > 0.05). While the mean skin temperatures in different groups were significantly diffe

the heat exchange between the trunk and the thermal environmentduring the exposure. The significant variation in the local skintemperatures on the extremities indicates that human foot, armand hand are more sensitive to thermal environment.

In Zhang’s study on the relation between local and overallthermal comfort [25], human body parts were grouped into threegroups in terms of the influence on the overall sensation: the mostinfluential group (chest, back and pelvis), the least influential group(hand and foot) and the moderately influential group (head, face,neck, breathing zone, upper and lower arms, thigh, lower leg). Thelocal sensation from the most influential group directly impacts theoverall sensation, though the local skin temperatures change rela-tively little (see also Fig. 6). Therefore, when selecting an existingMST calculation method or establishing a new one for the use instudies on thermal comfort, the skinmeasurement sites in themostinfluential group are necessary to reflect the overall thermal

increasing.

of mean skin temperature (�C)

2 3 4

32.91 � 0.2133.27 � 0.2133.63 � 0.17

32.68 � 0.1933.13 � 0.20

33.65 � 0.17

32.66 � 0.1733.29 � 0.2233.69 � 0.20

32.73 � 0.1533.25 � 0.1733.63 � 0.19

32.61 � 0.1633.15 � 0.20

33.59 � 0.18

e mean skin temperatures in the same group were insignificantly different (n ¼ 11,rent (n ¼ 11, P < 0.05).

Fig. 6. Local skin temperatures (Means) under different air temperatures. (a) Air temperature increasing, (b) Air temperature decreasing.

W. Liu et al. / Building and Environment 46 (2011) 478e488 487

sensation. On the other hand, in spite of the weak contribution tothe overall sensation, the least influential group plays a veryimportant role to adjust the whole body’s heat loss and satisfy thethermoregulation needs, due to the rapid and large variation in thelocal skin temperatures as depicted above. In order to reflect thechange of body’s thermal state well, the skin measurement sites inthe least influential group (including upper and lower arms, whichis in the moderately influential group) should also be seriouslyconsidered.

The effect of the order of air temperature changing on thevariation in local skin temperatures can also be found in Fig. 6. Itseems that the local skin temperatures at most measurement sites(except for forehead and trunk) had a larger change when the airtemperature was decreasing (see Fig. 6b) than when the airtemperature was increasing by the same degree (see Fig. 6a). Thismight be induced by the distinct impact of vasoconstriction (due toair temperature decreasing) and vasodilation (due to air tempera-ture increasing) on skin temperature. The trend implied a morerigorous requirement on sensitivity of an MST calculation method

Table 9Relation between thermal sensation and mean skin temperature.

Experimental condition Coefficient MST calculation method

4b 10c 17

Group 1 a 2.7928 3.1653 2.6165b �92.456 �104.570 �86.222R2 0.9108 0.9969 0.9652

Group 2 a 2.0170 1.8795 1.9709b �67.303 �62.819 �65.859R2 0.9358 0.9496 0.9795

Regression equation MSV ¼ a*MST þ b

MSV means mean sensation vote and MST mean skin temperature. R2 is determi-nation coefficient. For group 1 the environmental temperature changed from 21 to29 �C and for group 2 from 29 to 21 �C.

when used to reflect the change of MST with increasing airtemperature. The result of sensitivity evaluation suggested thatonly two of the five MST calculation methods evaluated as highreliability can meet the requirement. They are the calculationmethods 10c and 17. As listed in Table 1, the method 10c has fourmeasurement sites located on foot, hand (the least influentialgroup) and arm, and the method 17 arranges six measurements onthose parts, which warrant their high sensitivity. The other threemethods have less measurement sites on foot, hand, and arm (themethods 4b and 12) or smaller weighted factors for skin sites onextremities (the unweighted method 14).

The evaluation results of reliability and sensitivity revealed thatfor an MST calculation method high reliability does not necessarilyimply high sensitivity. That means not all accurate MST calculationmethods are appropriate to use in researches on thermal comfort,especially for the transient thermal condition. The index of sensi-tivity also needs to be carefully regarded. By doing so, it can beobtained that both the MST calculation methods 10c and 17 wereappropriate ones, with high reliability and sensitivity. If the index ofmeasuring sites number was taken into account, the method 10cwas more convenient and least time consuming due to its fewermeasuring sites. Therefore, it is recommended that the method 10cwas the most appropriate one based on the results in the presentstudy. As to the method 4b, though its sensitivity was not highenough, it was ideally suited to field studies when a large number ofsubjects are involved, because of using only four measuring sites.

The performances of the MST calculation methods to reflectthermal comfort were estimated, by relating the mean MST to themean sensation votes (MSV) under the four experimental temper-atures. According to the evaluation results depicted before, it issuggested that high reliability and sensitivity are the necessaryconditions for the MST calculation methods suitable to use inthermal comfort study. Therefore, only the recommended methods10c and 17, as well as the simplemethod 4b, were investigated here.

W. Liu et al. / Building and Environment 46 (2011) 478e488488

Table 9 gives the linear regression results for these calculationmethods. As shown in Table 9, a good linear relationship wasestablished between MSV and MST with a high determinationcoefficient (R2 > 0.9), when using any of the three MST calculationmethods. However, compared with the method 4b, the methods10c and 17 exhibited a better performance in reflecting thermalsensation, due to the bigger value of R2. This can be explained bybetter reliability and sensitivity of both methods. Another reasonmay be that the method 10c places three measurement sites in themost influential group of body parts and the method 17 has fivesites located in that group, while the method 4b arranges only onemeasurement site.

In our published work [26], fourteen MST calculation methodswere evaluated for use in predicting thermal comfort based on thePMVmodel. As indicated in that study, the methods 4b and 10c alsoobtained better results when applied in the PMV model at itsexperimental temperatures between 21 and 29 �C (the method 17was not investigated) [26].

The experiment in the present study was made under thenormal air temperatures (21e29 �C), which was not deviated verymuch from neutral environment. According to Olesen’s research[27], in warm conditions when skin temperature is rather uniform,2e4 measurement sites may be enough, while in cold conditions8e12 skin measurement sites may be necessary. The MST calcula-tion methods suggested in the present work possess at least 10sites. Therefore, it seems that the application of this study may notbe confined to within the given experimental condition. However,the effects of some extreme thermal environments (i.e. very coldand very hot) on the application of the MST calculation methodsstill need be investigated in future.

The following limitation may apply to the present study. Thenumber of the subjects was not large. Further studies need becarried out on the possible effect of gender difference involvingmore subjects and experimental conditions, in spite of the insig-nificant gender difference in this experiment. Though we endeav-ored to evaluate many MST calculation methods, some were likelyomitted because of our finite information. However, the evaluationmethod proposed in this study is almost available for various MSTcalculation methods.

5. Conclusion

Three indices including reliability, sensitivity and number ofmeasuring siteswereproposed to evaluateMSTcalculationmethodsfor the application in thermal comfort study. The result indicatesthat not all accurate MST calculation methods are ideal ones toreflect thermal comfort. Their sensitivity also needs to be seriouslyconsidered. Among the MST calculation methods evaluated in thisstudy, the methods 4b, 10c, 12, 14 and 17 had high reliability, whileonly the methods 10c and 17 held enough high sensitivity. Consid-ering the number ofmeasuring sites, it is suggested that themethod10c was most suitable for use in thermal comfort study. Theperformances to reflect thermal comfort were good, when the rec-ommended MST calculation methods were applied.

Acknowledgements

The project is financially supported by the Key Program ofNational Natural Science Foundation of China (Grant No.50838009).

The authors would like to acknowledge Qi Shen and ChunhongWang for their assistance during the experiment and the subjectswho volunteered for this study. And also, the authors want toexpress thanks toMsAiling LianandMsLi Lan for their improvementof the paper.

References

[1] ASHRAE. Thermal comfort. ASHRAE handbook: fundamentals. Atlanta:American Society of Heating, Refrigerating and Air-Conditioning Engineers,Inc; 2001.

[2] Mayer E. Objective criteria for thermal comfort. Building and Environment1993;28:399e403.

[3] Fanger PO. Thermal comfort. New York: McGraw-Hill Book Company; 1972.[4] Hoppe P. Different aspects of assessing indoor and outdoor thermal comfort.

Energy and Buildings 2002;34:661e5.[5] Bulcao CF, Frank SM, Raja SN, Tran KM, Goldstein DS. Relative contribution of

core and skin temperatures to thermal comfort in humans. Journal of ThermalBiology 2000;25:147e50.

[6] Gagge AP, Stolwijk AJ, Salting B. Comfort and thermal sensations and associ-ated physiological responses during exercise at various ambient tempera-tures. Environment Research 1969;2:209e29.

[7] Hasebe Y, Iriki M, Takahasi K. Usefulness of ReR interval and its variability inevaluation of thermal comfort. International Journal of Biometeorol1995;38:116e21.

[8] Huizenga C, Zhang H, Arens E, Wang D. Skin and core temperature response topartial- and whole-body heating and cooling. Journal of Thermal Biology2004;29:549e58.

[9] de Dear RJ, Ring JW, Fanger PO. Thermal sensations resulting from suddenambient temperature changes. Indoor Air 1993;3:181e92.

[10] Zhao R. Investigation of transient thermal environments. Building and Envi-ronment 2007;42:3926e32.

[11] Elmar M, Jan L, Gerd N. Heart rate variability a noninvasive approach tomeasure stress in calves and cows. Physiology & Behavior 2002;75:251e9.

[12] Kataoka Y, Yoshida F. The change of hemodynamics and heart rate variabilityon bathing by the gap of water temperature. Biomedicine & Pharmacotherapy2005;59:S92e9.

[13] Kleiger RE, Miller JP, Bigger JTJ, Moss AJ. Decreased heart rate variability andits association with increased mortality after acute myocardial infarction.American Journal of Cardiology 1987;59:256e62.

[14] Kleiger RE, Bigger JT, Bosner MS, Chung MK, Cook JR, et al. Stability over timeof variables measuring heart rate variability in normal subjects. AmericanJournal of Cardiology 1991;68:626e30.

[15] Malik M, Farrell T, Cripps T, Camm AJ. Heart rate variability in relation toprognosis after myocardial infarction: selection of optimal processing tech-niques. European Heart Journal 1992;10:1060e74.

[16] Myers G, Workman M, Birkett C, Ferguson D, Kienzle M. Problems inmeasuring heart rate variability of patients with congestive heart failure.Journal of Electrocardiology 1992;25:214e9.

[17] Lu G. Medical neurobiology. Beijing: Higher Education Press; 2000 [inChinese].

[18] Liu W, Lian Z, Liu Y. Heart rate variability at different thermal comfort levels.European Journal of Applied Physiology 2008;103:361e436.

[19] Mitchell D, Wyndham CH. Comparison of weighting formula for calculatingmean skin temperature. Journal of Applied Physiology 1969;26:616e22.

[20] Choi JK, Miki K, Sagawa S. Evaluation of mean skin temperature formulas byinfrared thermography. International Journal of Biometeorol 1997;41:68e75.

[21] Lee HM, Cho CK, Yun MH, Lee MW. Development of a temperature controlprocedure for a room air-conditioner using the concept of just noticeabledifference (JND) in thermal sensation. International Journal of IndustrialErgonomics 1998;22:207e16.

[22] ISO 9886. Evaluation of thermal strain by physiological measurements.Geneva: International Organization for Standardization; 1992.

[23] Ouyang H. Clothes hygiene. Beijing: People’s Military Medicine Press; 1985 [inChinese].

[24] ISO 9920. Thermal environment e estimation of the thermal characteristics ofa clothing ensemble. Geneva: International Organization for Standardization;1992.

[25] Zhang H. Human thermal sensation and comfort in transient and non-uniform thermal environments (PhD thesis). University of California, Ber-keley, 2003.

[26] Yao Y, Lian Z, Liu W, Jiang C. Measurement methods of mean skin tempera-tures for the PMV model. HVAC&R Research 2008;14(2):161e74.

[27] Olesen BW. How many sites are necessary to estimate a mean skin temper-ature. New York: Raven Press; 1984.