CLIMATE RESEARCHClim Res

Vol. 24: 267–284, 2003 Published September 19

1. INTRODUCTION

Clothing can be looked at from a variety of perspec-tives. It can be described in physical terms with respectto thermal resistance or insulation (Gagge et al. 1941,ASHRAE 1997) and its impedance to mass (watervapour) transfer (Woodcock 1962, Goldman 1981).From an ergonomist’s perspective, clothing can be re-garded as a mechanism capable of making a workplaceenvironment safer and healthier for workers, or in somecases, present a workplace hazard. From an anthropo-logical perspective clothing represents a significantcultural development that has, along with shelter and

fire, allowed the species to venture well outside its orig-inal tropical domain (McIntyre 1980, Clark & Edholm1985, Parsons 1993). From a social angle, clothing canbe variously thought of as a projection of personality,mood, religion, sub-cult and other group affiliations.Clothing also represents a code for organisational orcorporate identity, and of course it is very commonlyused to convey socio-economic status cues. But inconventional thermal comfort theory, perhaps mosteloquently described by Fanger in 1970, clothing sim-ply represents a single layer of thermal insulation uni-formly interposed between the human subject’s bodysurface and their immediate thermal environment.

© Inter-Research 2003 · www.int-res.com*Corresponding author. Email: rdedear@laurel.ocs.mq.edu.au

Weather, clothing and thermal adaptation toindoor climate

Craig Morgan, Richard de Dear*

Division of Environmental and Life Sciences, Macquarie University, Sydney, New South Wales 2109, Australia

ABSTRACT: The adaptive thermal comfort model links indoor comfort temperatures to prevailingweather outdoors, shifting them higher in warm weather and lower in cool weather. Adaptive com-fort engineering standards hold the potential to conserve energy, but for them to work effectively itis essential that building occupants are free to adapt themselves, primarily through clothing adjust-ment, to the variable indoor climatic regimes prevailing inside such buildings. This paper examinesclothing behaviour and its relationship with thermal environments in 2 different indoor settingslocated in Sydney, Australia. The first was in a suburban shopping mall, and the second in a call-centre office. The company operating the call-centre had a strict business attire dress-code in forceMondays through Thursdays, but on Fridays employees were free to wear casual clothes. Indoor tem-peratures throughout both studies were relatively static despite significant weather and seasonaltrends operating in the outdoor atmospheric environment. Clothing insulation values of garmentsworn inside the shopping mall showed significant day-to-day variation, with the standard deviationrepresenting about 34% of the mean clo (1 clo = 0.155 m2 K W−1) value during the study period. Typ-ical shopping mall clo values during summer were more than 0.6 clo lighter than those worn in win-ter, and a regression model of daily average clo on daily mean outdoor temperature explained 52%of the variance in clo values. Clothing insulation levels worn on ‘strict business attire’ days in the call-centre study were relatively static, regardless of weather or season, and they showed only a weak sta-tistical relationship with indoor temperature variations (R2 = 20%). In contrast, the much more vari-able clo values worn by the office workers on ‘casual’ days, showed a significant correlation withoutdoor temperatures (R2 = 44%). Based on these relationships, an adaptive comfort model for use inhybrid ventilation buildings is proposed.

KEY WORDS: Thermal comfort · Hybrid insulation · Clothing insulation · Adaptive comfort model ·Indoor climate · Energy conservation

Resale or republication not permitted without written consent of the publisher

Clim Res 24: 267–284, 2003

The significance of this heat-balance parameter hasnot escaped the attention of adaptive comfort modelproponents, who advocate wider indoor temperatureranges than conventional comfort standards and prac-tices currently permit, with a view to conservingenergy in the built environment (Humphreys 1979,Auliciems 1986, de Dear & Brager 1998, 2001). Empiri-cal evidence of the role played by clothing insulation inadapting to indoor climate can be seen in Fig. 1. Takenfrom de Dear & Brager’s global intercomparison ofthermal comfort field studies, the graph indicates thatthe mean thermal insulation1 worn inside any singlebuilding (mostly offices) tends to correlate negativelywith average temperatures inside the building (deDear & Brager 1998, 2001). Based on some 22 000 ques-tionnaire responses across 160 different buildings scat-tered across 4 continents, the regression relationshipdepicted in Fig. 1 accounts for about a quarter of thetotal between-building variance in average indoorclothing insulation estimates, suggesting that indoortemperature is an important determinant of clothingbehaviour indoors but also that there are many otherinfluences as well as some random variability.

The field evidence reported above is further rein-forced by recent longitudinal observational studies

that indicate building occupants modify their clothingseveral times throughout the day, largely in responseto non-steady-state thermal comfort conditions indoors(Baker & Standeven 1996, Newsham & Tiller 1997). Onaverage, 15% of the occupants in their Canadian officebuildings modified their clothing during the hour priorto interview. Even minor adjustments such as loosen-ing a neck-tie or rolling up shirt-sleeves, which wouldbe barely measurable using contemporary thermalmanikins, should be regarded as purposive thermoreg-ulatory behaviour.

Another feature of Fig. 1 is the remarkable differ-ence in clothing variability across different buildings,as indicated by the error bars (±standard deviation)around each building’s mean clo value. At one extremethere are buildings where the central 2 standard devi-ations in clo span a range of 1.0 clo, while at the otherextreme the within-building heterogeneity is less than0.1 clo. The reasons for such diversity in within-build-ing clo variance are many and varied, some biophysi-cal and others social. An obvious biophysical hypothe-sis is the degree of indoor climatic homogeneity. Themore spatially and temporally diverse the indoor con-ditions within a building, the more variable the cloth-ing responses by that building’s occupants. Anotherfactor explaining diversity of within-building clo vari-ances is the overall level of warmth prevailing insidethe building. Fig. 1 demonstrates converging insula-tion values, both between and within buildings, asindoor temperatures increase. This suggests that thedegrees of freedom for clothing behavioural ther-moregulation diminish as the number of garmentsdecreases towards a socially acceptable minimumlevel, ca. 0.45 clo (or 0.6 clo in Fig. 1, which includes anaverage insulation increment of 0.15 clo for the officechair being used at the time of questionnaire).

As noted earlier, clothing insulation levels are notdetermined solely by indoor climatic factors. Diversityamong the occupants themselves could also contributeto differing within-building clo variances. For exam-ple, a building with exclusively same-sex occupantscould be expected to have greater homogeneity in clovalues than one with both females and males. Therehas long been recognition of a difference between theinsulation values worn by males and females occupy-ing the same office, with women wearing significantlyless than males in summer, and more in winter (e.g. deDear & Brager 1998, 2002). There is also an organisa-tional factor contributing to the diversity of clo vari-ances depicted in Fig. 1. Nearly all the buildings in thesample were commercial or governmental offices, andthese organizations often have dress codes regulatingwhat their employees can and cannot wear in theworkplace. By prescribing strict clothing standards oreven uniforms, these more formal dress codes carry

268

1One clo approximates the thermal insulation value of aheavy winter business suit, or 0.155 m2 K W−1

Fig. 1. The relationship between clothing insulation levelsworn inside buildings (naturally ventilated and air-condi-tioned) and the mean indoor temperatures in those buildings.The data are from the ASHRAE global thermal comfort data-base (de Dear 1998). Each point represents a building aver-age clothing insulation, as estimated by a standardised gar-ment checklist and calibrated against the methods set out inASHRAE Standard 55 (1992). Indoor temperatures are basedon the operative temperature index, which is an arithmeticaverage of air and mean radiant temperatures. Error bars rep-resent ±1 standard deviation of the building sample’s clovalue estimates. Clo estimates include the incremental effectof the upholstered chair upon which the subject was sitting at

the time of the questionnaire (ca. 0.15 clo)

Morgan & de Dear: Weather, clothing and indoor climate

significant implications for the adaptive model of ther-mal comfort—namely, they limit thermal adaptiveopportunities for employees in office workplaces.

Indoor clothing insulation levels also reflect outdoorenvironmental conditions. The adaptive comfort modelprotagonists have long argued that building occupantsrespond not only to the thermal regimes insidetheir buildings but to the weather and seasonal condi-tions prevailing outdoors as well (Auliciems 1981,Humphreys 1981, de Dear & Brager 2001). This hypoth-esis arises from the common-sense observation thatweather forecasts are routinely issued during the morn-ing, at a time when they can affect clothing-wardrobedecisions. But of course another clothing−weather link-age is the fact that thermal regimes inside certain typesof buildings are actually correlated to varying extentswith temperatures prevailing outside the building. Thisis particularly so in the case of ‘free-running’ (neitherheated nor cooled) buildings, or those with natural ven-tilation. Fig. 2 depicts the relationship between themean thermal insulation worn inside buildings in thede Dear & Brager (1998, 2001) study and the mean out-door effective temperature (ET*) at the time of thestudy. The strength of the relationship is summarisedby the R2 statistic (coefficient of determination), whichindicates that about half (49%) of the between-buildingvariance in clo values was explained by variations inoutdoor atmospheric conditions.

Despite these direct and indirect linkages betweenclothing on the one hand and both indoor and outdoorclimates on the other, the thermally regulable dynam-ics of clothing behaviour are often overlooked intoday’s thermostatically regulated workplaces. An-thropologists would indicate that clothing, as a part of

material culture, originally evolved to insulate our-selves from thermal situations that we find unaccept-able or indeed stressful (Parsons 1993), but in contem-porary urban societies, where it is typical to spendmore than 90% of our lives indoors (Fanger 1970), theresponsibility of thermoregulation seems to lie nowwholly with air-conditioning systems and not the indi-vidual building occupant (Fountain et al. 1996). Thisshift in ‘thermal comfort responsibility’ adds furtherimportance to the regulation and development ofindoor climate standards such as ASHRAE Std 55-92(ASHRAE 1992). It also diminishes the thermoregula-tory significance of clothing, allowing the non-thermalfactors identified earlier to gain ascendancy in clothingdecisions. For example, since indoor climates can beengineered to suit any arbitrarily chosen clothing insu-lation level, office-based workplaces are now free toenforce whatever style and quantity of clothing theydeem appropriate to their corporate image. The endresult has been a globalisation of indoor clothing pat-terns, with office workers in the humid tropics ofSoutheast Asia, for example, wearing very similarbusiness suits to their counterparts in temperate oreven subarctic climate zones. We can, and do, wearwhatever we like indoors in virtually every climatezone on the planet, providing we are prepared to paythe financial and environmental costs of heating, ven-tilation and air-conditioning (HVAC) energy (Fountainet al. 1996).

However, it would be wrong to construe contempo-rary clothing behaviour as totally devoid of any ratio-nal basis (Clark & Edholm 1985). Markee White (1986)investigated the various factors impinging upon cloth-ing garment and ensemble choices of office workers inNorth America. Using a questionnaire it was noted thatthe salient factors were, in order of importance:(1) comfort (non-thermal wear factors), (2) anticipatedoutdoor environment, (3) anticipated indoor environ-ment, (4) appropriateness for the job, (5) desire tobe fashionable, (6) dress code, and (7) after-workactivities.

The same subjects said that they could have dressedmore comfortably for work were it not for (in order ofimportance): (1) inappropriateness for the job, (2) dresscode of their workplace, (3) inability to anticipateindoor conditions, (4) desire to be fashionable, and(5) lack of more comfortable clothing options in theirwardrobe. The first 2 factors in this list appear to berelated—the first one is like an ‘unwritten’ dress code,whereas the second is a clearly articulated policy onwhat is permissible clothing.

Five research hypotheses arise from this literaturereview of clothing, thermal comfort, and indoor- andoutdoor climate, and these are the foci of the researchreported here:

269

Fig. 2. Relationship between the mean thermal insulationworn inside buildings in the de Dear & Brager (1998; 2001)study and the mean outdoor effective temperature (ET*) atthe time of the study. The latter was calculated as the averageof daily minimum ET* (air temperature, relative humidity and

clo values as input) and daily maximum ET*

Clim Res 24: 267–284, 2003

• (1) There are gender differences in both levels anddynamics of clothing insulation worn indoors.

• (2) Clothing thermal insulation levels worn indoorsare affected by context/setting (for example, officevs department store).

• (3) Clothing insulation levels worn indoors areaffected by corporate dress codes (for example, thedistinction between formal office attire and casualclothing).

• (4) Clothing insulation levels worn indoors areaffected by the indoor climate variability.

• (5) Clothing insulation levels worn indoors areaffected by the outdoor weather and climatic contextof the building.

2. METHODS

We investigated these research issues by observingthe clothing behaviour of subjects within 2 distinctresearch designs. First we used a cross-sectional studyof patrons in a large suburban shopping mall. Theseobservations were supplemented by a longitudinalstudy of a smaller sample of office workers in a modernopen-plan office (a telemarketing call-centre). Theparticular buildings in which we performed the studieswere selected on the basis of permission being grantedby employers and tenants. The cross-sectional studywas used to investigate average clothing insulationlevels across a large sample (about 45 new subjects perday) every day for about 6 mo. The second study, lon-gitudinal in design and of 5 mo duration, was used totrack clothing behaviour in relation to indoor and out-door climatic variations and to assess the impact ofdress codes on adaptive opportunities.

2.1. Shopping mall study

The first study was conducted between August 1996and January 1997 in a department store located withina shopping mall in the western suburbs of Sydney,Australia (latitude 34° S). Every fourth adult subject topass the observation station within the departmentstore was unobtrusively observed by the researchers(Moser & Kalton 1971), and their clothing insulationvalue rated in units of clo. The method of clo estimationwas based on the garment check-list defined inASHRAE Std 55-92 (ASHRAE 1992). The cumulativetotal of subjects’ clothing garment effective insulationvalues was registered as the intrinsic insulation valueof the composite ensemble. Table 1 lists the effectivegarment insulation values used in this study.

(1)

where Icl is the intrinsic insulation value of the com-posite ensemble and Iclu,i is the effective insulationvalue of the i th garment.

Observations were made on every day of the 6 moperiod, providing an uninterrupted time series for sta-tistical analysis. Observers were trained for 2 d on-sitebefore the data reported in this paper were collected.Concurrent outdoor temperatures were obtained froman automated weather station less than 1 km from theshopping mall site. Indoor temperatures were recordedwith a linearised composite thermistor connected to a

I I i

icl clu= ∑ ,

270

Garment description Iclu,i (clo)

Bra 0.01Panties 0.03Full slip 0.16Half slip 0.14Men’s briefs 0.04T-shirt thin 0.08Ankle socks 0.02Calf socks 0.03Knee socks (thick) 0.06Pantyhose/stockings 0.02Shoes 0.02Boots 0.10Sandals/thongs 0.02Sleeveless blouse 0.13Short-sleeve blouse 0.19Short-sleeve knit sport shirt 0.17Long-sleeve shirt 0.25Long-sleeve flannelette shirt 0.34Short shorts 0.06Walking shorts 0.08Short-sleeve T-shirt 0.08Thin trousers 0.15Thick trousers 0.24Sweatpants 0.28Overalls 0.30Thin suit vest 0.10Thick suit vest 0.17Thin single breasted jacket 0.36Thick single breasted jacket 0.42Thin double breasted jacket 0.44Thick double breasted jacket 0.48Thin sleeveless vest (sweater) 0.13Thick sleeveless vest (sweater) 0.22Thin long sleeve sweater 0.25Thick long sleeve sweater 0.36Thin skirt 0.14Thick skirt 0.23Sleeveless dress thin 0.23Short-sleeve dress thin 0.29Long-sleeve dress thin 0.33Long-sleeve dress thick 0.47Long-sleeve T-shirt thick 0.20

Table 1. Individual clothing garments and their effective insu-lation values (Iclu,i, clo). Ensemble intrinsic insulation valuesare derived by summing individual garment effective insula-

tion values (after ASHRAE Standard 55-92)

Morgan & de Dear: Weather, clothing and indoor climate

portable datalogger with a logging interval of 10 min.Typically there were about 45 different subjectsassessed on any one day, providing the basis for a dailymean clothing insulation value.

2.2. Office building study

The second study was conducted in a typical ‘Grade-A’ air-conditioned office building in the central busi-ness district of Sydney, Australia. Between September1997 and January 1998 a sample of 14 call-centreworkers (7 female and 7 male) were studied in relationto their clothing insulation levels and thermal comfort.Their average age was approximately 35 yr. This officeenvironment was fundamentally different to the shop-ping mall study because dress codes were in force.From Monday through Thursday of each week therewas a strict business attire dress code in place, necessi-tating suits and other formal clothing. However on Fri-days, the employer had a policy of relaxing the dresscode and permitting staff to wear casual clothing (typ-ically jeans, shorts and T-shirts).

The ScreenSurvey™ software (Newsham & Tiller1997) was used to gather data from a sample of 14 sub-jects. The software designed and presented question-naires in the familiar Windows™ interface. It automat-ically ‘popped up’ on the subjects’ workstation screensat least once each day. When the questionnairepopped up, the subject was required to go through aclothing garment checklist (see Table 1) and also tocomplete a small battery of thermal comfort questions(not reported in this paper). The whole questionnairerequired less than 2 min to complete, but if the subjectwas busy when the questionnaire popped up, theycould postpone it for up to an hour. The data weredownloaded from the subjects’ computers at the end ofeach working week and imported to our database forsubsequent statistical analysis. On any one poll (everyhalf-day) we registered responses from an average of10 out of the 14 subjects.

Before reaching the clothing section of the software,subjects were required to indicate gender, and thisresponse was used by the software’s algorithm to pre-sent only the relevant half of the garment checklistfrom Table 1. The effective insulation values of eachgarment (Iclu,i) in the questionnaire checklist werecoded into a look-up table (hidden from subject’sview), and the final ensemble intrinsic insulation (Icl)was automatically time-stamped and logged to thedata-export file.

Local and concurrent outdoor meteorological datawere obtained from the Australian Bureau of Meteo-rology’s automated weather station (AWS) at Observa-tory Hill, within the Sydney central business district

and less than 2 km from the office building underobservation. The AWS temperature and humidity datawere sorted into 15 min bins. All temperatures for the15 min data periods within the building’s office hourseach day were averaged to yield a daily mean temper-ature.

Indoor temperatures for the call-centre were re-corded by a pair of linearised composite thermistors,each connected to its own miniature datalogger. Theywere placed within the occupied zone of the call-cen-tre. One was located on a structural pillar in the build-ing’s HVAC core zone, approximately 10 m from theexternal glazing. The other was within the zone wherethe workstation was located, 0.6 m above floor level.Local thermal effects produced by computers andother heat-emitting office equipment, supply air vents,and open doorways were taken into considerationwhen placing the temperature loggers.

3. RESULTS

The project’s results are organised according theenvironment in which they were collected.

3.1. Shopping mall results

The outdoor weather observations during the shop-ping mall study are plotted in Fig. 3. They show theexpected gradual increase in air temperatures from theaustral spring into summer. The highest daily averageoutdoor temperature was recorded during a brief heat-wave centred on 11 October, quite early in the australspring. Other warm-to-hot spells with mean daily tem-peratures exceeding 25°C occurred on 14 Novemberand again on 2 December. The overall mildness of Syd-ney’s climate is evident in Fig. 3, which indicates meandaily temperatures above 10°C (July) and below 25°C(January). After ranking the mean daily temperaturesfor the entire 6 mo duration of the study we found the10th percentile was 12.5°C, the 90th percentile was22.1°C, while the median (50th percentile) was 17.3°C.This median temperature came within half a degree ofthe long-term climatological mean for western Sydney(Bureau of Meteorology 1991), suggesting that the6 mo study period was climatically typical for theregion. Despite the minor seasonal contrasts in temper-ature for this part of the world, the day-to-day standarddeviation in mean outdoor temperature of 3.8°C sug-gests a fairly changeable synoptic environment duringthe 6 mo study period.

Indoor air temperatures within the shopping mall arealso plotted in Fig. 3. The graph indicates a veryrestricted range of indoor temperatures throughout the

271

Clim Res 24: 267–284, 2003

6 mo study, despite significant external air tempera-ture variability. Indoor temperatures were generallycontained within the 22 to 24°C band. The observationof just 3 indoor temperature excursions above 25°Cthroughout the entire 6 mo, two of which fell in wintermonths, underscores the absence of any seasonal ad-aptation in the shopping mall’s HVAC system.

Overall, the clothing style observed in the mall studycould best be classified as ‘casual’. The daily clo valuesplotted in Fig. 4 are averaged across all subjectsassessed on a given day (sample size ca. 45). Dailymean clo values trend downwards from winter through

spring and into summer. Apart from theanomalously low-clo week at the start of thestudy, mean wintertime clothing insulationvalues were generally in the 0.7−0.9 clorange, dropping to the 0.3−0.5 clo range bymid-summer (January).

Fig. 5a shows the scatterplot of daily meanindoor temperature and daily mean clo val-ues. The statistically insignificant regressionmodel (F = 0.77, p = 0.37, df = 1,171), with anegligible coefficient of determination (R2 =0.0045), suggests that clothing insulation lev-els worn inside the department store were in-dependent of temperatures prevailing withinthe store. Fig. 5b is a regression of mean dailyclo observations on mean daily outdoor tem-perature, as recorded at the nearby automaticweather station. The regression model washighly significant (F = 188.4, p < 0.0001, df =

1,171), with 52% of the day-to-day variance in meandaily clo values being accounted for by the relationshipwith daily outdoor temperature. Of the various regres-sion function options we found that a power functionmaximised the explained variance (R2) in this case.Fig. 5b indicates a gradual decay of indoor clo valuesfrom ~0.9 on the coldest days (mean outdoor tempera-tures below 10°C) towards a minimum daily averageclo of 0.2−0.3 clo on the warmest days (mean outdoortemperatures above 25°C).

The statistical association between daily mean clo val-ues and outdoor temperatures was extended beyond the

272

Fig. 3. Indoor and outdoor daily average temperatures during the shop-ping mall study

Fig. 4. Daily averaged clothing insulation values in the shopping mall study

Morgan & de Dear: Weather, clothing and indoor climate

day on which the clo measurements were taken, to sev-eral days preceding. Pearson product-moment correla-tion coefficients were calculated between daily mean clovalue on Day x and daily mean outdoor air temperatureson Day x–1, Day x–2, Day x–3 … Day x–7. Fig. 6 showsthe strength of correlation (expressed as R2). A gradualdecay in correlation between Day x clo and outdoor tem-perature is evident as the time-lag increases from 0 to7 d. An exponential regression model fitted to the seriesof 8 time-lagged coefficients of determination (R2) wasable to explain 97% of the variance. All 8 regression co-efficients were statistically significant at better than 0.05probability. In contrast there was no apparent relation-ship, statistical or otherwise, between daily mean clo val-ues and indoor temperatures, regardless of which time-lag was applied.

Gender represents another dimension in the clothingmatrix that can be examined with this data set. Whileno questionnaire was actually administered to subjectsin the shopping mall study, our field researchers wererequired to record gender to enable the correct clo-value look-up table (Table 1) to be used when subse-quently estimating garment and ensemble insulationvalues.

Table 2 indicates that female subjects wore, on aver-age, marginally warmer clothing than the males. Butperhaps more important to the focus of this paper is theobservation that females registered greater variabilityin their daily mean clo values than the males. This canbe seen in Fig. 7, with both genders’ regression modelsbeing statistically significant at better than 0.05 proba-bility.

273

Fig. 5. Regression between mean daily clothing and mean daily (a) indoor and (b) outdoor temperatures during the shopping mall study

Fig. 6. Coefficients of determination (R2) for the relationship between mean daily indoor clothing insulation and indoor/outdooraverage temperatures across various time-lags (from 0 to 8 d). All 8 coefficients were statistically significant at better than the

0.05 level

a b

Clim Res 24: 267–284, 2003

3.2. Office environment results

Fig. 8 shows the gradual upward trend of mean dailytemperatures from about 15°C in October up to themid-20s in January during the office environmentstudy. The office building was largely unoccupied dur-ing the weekends, hence the weekly truncations in thetime-series in Fig. 8.

Fig. 9 shows indoor the temperature observationsduring the office environment study. As was the casewith the shopping mall study, there was a relativelyhomogeneous mean daily indoor temperaturethroughout the study period. However, the actual tem-peratures recorded in the office, in the mid-20s, weregenerally a few degrees higher than those recorded inthe shopping mall study. Fig. 9 indicates a significantamount of temperature variation within each workingday at the call-centre, with ranges of 5 to 6ºC beingtypical during office hours.

Intrinsic ensemble thermal insulation values werecalculated and a daily averaged clo figure determinedby the on-screen questionnaire software automaticallyafter the subject had completed the garment check-list. The subjects’ employer had a strict ‘business attire’dress code policy on the days between Monday andThursday. On Fridays, conversely, the dress code was‘casual’. Fig. 10 shows the plot of all clothing insulationresults, with Fridays being last data point in each string

274

Female Male

Mean ensemble insulation 0.51 0.47values

Standard deviation of daily mean 0.18 0.15insulation values

Table 2. Mean ensemble insulation values (clo) observed in the shopping mall study broken down by gender

Fig. 7. Regressions between (a) male and (b) female mean daily clothing insulation values and mean daily outdoor temperatures during the shopping mall study

Fig. 8. Average daily outdoor air temperature during the office environment study, with daily minima and maxima superimposed

a b

Morgan & de Dear: Weather, clothing and indoor climate

of 5. The daily averaged (across all subjects) clo valuesobserved on Friday’s were noticeably higher thanthe other weekdays’ values in winter, and lower insummer.

The statistical relationships between average dailyclothing insulation being worn indoors and mean dailytemperature, both indoors and out, were assessed byregression analysis. R2 for the indoor office-hours tem-perature model is negligible at 0.0026 (F = 0.2, p > 0.05,

df = 1,79). The regression model between daily meanoutdoor temperature and all clothing data has a signif-icant R2, 0.143, but it is also relatively small (F = 12.9,p < 0.05, df = 1,79).

The clear differentiation of clothing styles into‘business’ (Monday through Thursday) or ‘casual’(Friday) attire invites another way of looking at thedata. Fig. 11 shows the relationship between meandaily indoor clothing data identified as business attire

275

Fig. 9. Mean temperatures indoors during office hours, with daily minimum and maximum superimposed, during the office environment study

Fig. 10. Daily averaged clothing insulation during the office environment study

Clim Res 24: 267–284, 2003

and both indoor and outdoor daily average tempera-ture. The binary group-average daily clo valuesrecorded on business attire days show negligibledependence on either indoor or outdoor tempera-tures.

Insulation values of clothing ensembles worn on‘casual Fridays’ (i.e. mufti2 days) are plotted in relationto indoor and outdoor temperature in Fig. 12. Despitethe apparent positive relationship between mean dailyoffice-hours indoor temperature and mean casualclothing insulation estimates in Fig. 12a, the small R2

(0.10) was statistically insignificant (F = 1.2, p > 0.05, df1,16). In contrast, 44% of the variance between dailymean clo values on casual Fridays was accounted forby the regression relationship with mean daily outdoortemperature (Fig. 12b; R2 = 0.44, F = 11.26, p < 0.01, df1,16).

4. DISCUSSION

4.1. Data-accuracy issues

Chapter 8 of the ASHRAE handbook of fundamen-tals (ASHRAE 1993) indicates that the most accuratemethods for determination of clothing insulation are(1) measurement on thermal manikins (McCullough &Jones 1984, Olesen & Nielsen 1983) or (2) direct mea-surements of skin and clothing surface temperaturesalong with clothing-layer heat fluxes ((Nishi et al.1975, Danielsson 1993). Estimation of ensemble insula-tion values from checklists of published garment val-ues is suggested in the ASHRAE handbook of funda-mentals to have an accuracy of the order of ±25% ofthe benchmark thermal manikin measurements(ASHRAE 1993). Explanations for the discrepanciesare legion, but uppermost would be the differences infabric weave, materials and fabric weight, not to men-tion variations in garment and ensemble fit which, inturn, modify the amount of sub-clothing layer air avail-able to thermally insulate the subject. The on-screen

276

Fig. 11. Regression of Monday−Thursday (business attire) clothing insulation estimates on (a) indoor and (b) outdoor mean temperatures. Data from the office environment study

Fig. 12. Regression of casual Fridays (mufti day) clothing insulation estimates on (a) indoor and (b) outdoor mean temperatures. Data from the office environment study

2The word mufti (pronounced \muf’ti\,) was once used to de-scribe civilian clothes when worn by a naval or military offi-cer; a term derived from the British service in India

a

a

b

b

Morgan & de Dear: Weather, clothing and indoor climate

clo-checklist method used in the office environmentstudy would be subject to the ±25% errors suggestedby ASHRAE. Such an error figure is far from ideal, butthe decision to use real office workers in a field settingprecluded any of the laboratory methods describedabove. Whatever errors exist in our raw data, they areunsystematic and uniformly distributed throughout thesample.

The method of clothing insulation estimation used inthe shopping mall study is probably subject to errorseven larger than those discussed above, because theresearchers were unable to actually speak to the sub-jects or get them to complete a garment checklist. Thetrained-observer method represents the only practica-ble solution to the need for high-speed unobtrusiveobservations in a real-world setting. A case could bemade that a trained observer would be more reliable atclassifying garments into the checklist than the sub-jects themselves, but there remains the problem ofunobserved garments or clothing layers hidden fromour observers’ view. Given the relatively mild climaticsetting of Sydney, we expect that the overwhelmingmajority of hidden garments worn by subjects in theshopping mall study would have been underwear andour assumption of standard underwear clo valueswould be realistic in at least 90% of cases. We base thisestimate on earlier Sydney thermal comfort studies inwhich large samples explicitly completed the standardASHRAE Std 55 garment checklists (Rowe 1996, deDear 1998). Another source of potential bias in the‘unobtrusive observer’ collection of clothing insulationestimates in our shopping mall study is the fact that thetrained observers were aware of the research hypothe-ses under investigation in this project. While this biasmay not manifest as explicit misclassification of gar-ment types (a short-sleeve shirt is a short-sleeve shirt,regardless of who observes it), the way in which adjec-tives such as ‘light, medium and heavy’ are applied toindividual garments could be biased by expectationson the part of the trained observers. Given the rela-tively mild climatic context of Sydney, the vast majorityof clothing garments in our database were classified as‘light’, and as such, this potential bias would be minor,if present at all.

So while there would inevitably be errors in the rawdata collected in this project we believe that they wererandomly and uniformly distributed throughout thesamples. In a colder climate zone than Sydney’s, thepotential for underestimating ensemble clo values dueto hidden garments or clothing layers in the shoppingmall study might become significant, particularly onvery cold days. However, in Sydney’s mild climatethere is no reason to suspect that such underestimationwould be large, or systematically correlated with themajor independent variables of the study. The focus of

this study was not so much to define the absolute valueof clothing insulation being worn by the populationsusing Sydney’s shopping mall and office environ-ments, but rather to assess the relationships betweenclo values worn indoors and both indoor and outdoortemperatures.

4.2. Gender issues and indoor clothing insulation

On average, females in the shopping mall studywore marginally heavier clothing than did the males inthe 6 mo study. Furthermore, their insulation valuesdemonstrated greater weather sensitivity than themales, as demonstrated by the females’ higherbetween-day variance (represented by R2) in mean clovalues that could be explained by variations in meandaily outdoor temperature in Fig. 7. Remembering thecontext in which these data were collected—namelycasually attired persons in a shopping mall—theheightened weather sensitivity for female clothinginsulation suggests a fundamental difference in cloth-ing behaviour between the two sexes. Casually attiredmales would appear to follow an informal dress codemore than women. The majority of males analysed inthe shopping mall study wore the ‘casual uniform’ ofjeans with either T-shirt or regular button-up shirt,regardless of weather dynamics outside. The ‘low-clo’option of shorts for the males was encountered in thisstudy very rarely, particularly for the ‘30 yr and over’age bracket. Women, on the other hand, were visiblymore weather responsive in their clothing choices. Forexample, light, short skirts/dresses or shorts and lightshort-sleeve tops were more prevalent in hot weatheracross all age brackets. Although the gender dimen-sion was not formally analysed in the office environ-ment study, the researchers formed the impression thatwomen had heightened weather sensitivity in theirclothing patterns than males in that context as well,especially on the casual Friday ‘mufti days’.

4.3. Contextual effects on clothing insulationworn indoors

A few basic questions arise from the comparisonsbetween the shopping mall and office settings in thisstudy. Firstly, there appears to be a more pronouncedseasonal variation in clo values in casual settings(shopping mall) compared to the more formal officesetting, but this is probably more the result of theoffice dress code dampening down the seasonalresponsiveness of clothing behaviour than any intrin-sic contextual effect. Another factor that couldexplain this observation is the duration of exposure.

277

Clim Res 24: 267–284, 2003

In the case of the subjects in a shopping mall, theymay expect to spend a reasonable part of their shop-ping trip actually exposed to outdoor weather condi-tions, and they may dress accordingly. In contrast theoffice workers may recognise that the duration oftheir exposure to office indoor climate will be moresignificant than the outdoor setting, and therefore theclothing insulation choices may reflect that relativeweighting of exposure durations. Thirdly, there ap-pears to be significantly greater inter-individual vari-ability in clo estimates in casual settings, suggestingthat inter-individual differences in thermal comfortpreferences are being accommodated by casual(unregulated) clothing patterns, where individualsare free to find their own thermal equilibrium bydressing however they wish. That is, given thechance, people will exercise the adaptive opportuni-ties (Baker & Standeven 1996) available to them inthe form of clothing choice and adjustments. A logicalextension of this line of reasoning is that there pre-sumably must have been more people enjoying ther-mal comfort (or, alternatively, fewer experiencingthermal discomfort) whenever and wherever they hadthe adaptive opportunity to behaviourally thermo-regulate their clothing.

4.4. Effects of indoor climate on clothing insulationlevels worn indoors

Indoor temperatures prevailing in the shopping mallstudy had no statistical association with clo levelsbeing worn inside the mall whatsoever. This can pre-sumably be put down to the very limited amount ofday-to-day variance in the department store’s indoortemperatures during the 6 mo study. Similarly, in theoffice building study there was a negligible statisticalrelationship between clo levels worn inside and con-current indoor temperatures. However, the extremelylimited variance in both the predictor (indoor tempera-tures) and the dependent (indoor clo) variables pre-cluded any other finding being made with the currentdata.

4.5. Effects of outdoor weather and climate on indoorclothing insulation

Visual analysis of the indoor clothing and outdoorweather time series in this study indicates a relation-ship between the 2 variables. The shopping mall cloth-ing insulation observations in Fig. 4 show a cleardownward trend in clo levels from winter (mid-year)into summer (end/start of year). The most prominentexception occurred towards the end of November (late

spring), when clo levels spiked upwards to levels notseen since the depths of winter, but this seems to beaccounted for by the sudden cold snap that occurred inSydney’s weather in that week (Fig. 3). There wereobvious variations in daily average clo values of about0.2−0.3 clo amplitude, with a periodicity spanning acouple of days—roughly corresponding to the synop-tic-scale variations evident in the outdoor temperaturetrace in Fig. 3. A more quantitative analysis of the rela-tionship between clothing and weather indicated thatmore than half of the day-to-day variance in mean(between-subjects) clo levels observed inside theshopping mall was accounted for by concurrent out-door daily temperatures (Fig. 5). However, it should beremembered that the present study’s data were col-lected from short-term visitors to a shopping mall. Itseems reasonable to expect that the relative signifi-cance of indoor versus outdoor temperatures on cloth-ing behaviour will be strongly influenced by the rela-tive duration of exposure to both atmosphericenvironments.

Compared to the shopping mall study, the officeclothing insulation data in Fig. 10 revealed a morecomplex relationship with outdoor temperature. Theclo levels observed in the office on Mondays throughThursdays were remarkably homogenous, with dailyaverages falling consistently between 0.7 and 0.8 clo.Even on 26–27 November, when a heat wave withmaximum temperatures reaching into the high 30soccurred, the formal business attire indoors remainedat the usual 0.7 clo. However, during Fridays, when theoffice workers were permitted to dress casually, therewas a much greater variability in thermal insulationbeing worn. Clo levels on casual Fridays during thecoolest month of the study (October; Fig. 10) were gen-erally heavier than during the other weekdays at thattime of year. But once the summer weather set in, thecasual Friday clothing insulation levels worn in theoffice were as much as 0.2 clo lighter compared toMondays through Thursdays.

The disconnection between formal business attireclo levels indoors and concurrent outdoor weather isvisually apparent in Fig. 11. In contrast, day-to-dayvariations in casual Friday clo levels were much moreclosely related to outdoor temperature (R2 = 0.44;Fig. 12b). This figure is possibly an underestimate ofthe true correlation because of the small numbers ofFridays falling within the study period compared to theother 4 days in the working week.

The evidence presented in this paper supports ourearlier findings that outdoor temperatures influencethe clothing levels worn indoors (de Dear & Brager1998, 2001). The effect seems most pronounced in situ-ations where people have some autonomy over whatthey wear, but just how this proposed cause-and-effect

278

Morgan & de Dear: Weather, clothing and indoor climate

linkage actually works is an interesting question initself. It is not difficult to understand how the tempera-ture of the indoor microclimate surrounding thehuman body exerts an influence on clo levels. Indoortemperature directly impacts the body’s heat balance,skin temperatures and skin wettedness, which are, inturn, the main thermophysiological drivers for thermaldiscomfort ((ASHRAE 1993, Gagge et al. 1986). In con-ventional thermal comfort theory we regard the moti-vation for clothing selection and indeed, any otherthermoregulatory behaviour, as being proportional tothe intensity of conscious sensations of thermal dis-comfort:

indoor climate ⇒ body heat balance ⇒ physiologicalstrain ⇒ thermal discomfort ⇒ behavioural thermo-regulation (clothing)

So if that’s the causal chain linking indoor tempera-tures to indoor clothing insulation levels, how can out-door temperature exert an effect as well, given that theindoor climate of both settings studied in this paperwere completely dissociated from outdoor tempera-tures (i.e. the buildings had centralized HVAC in oper-ation)? The answer may lie in the timing of exactlywhen clothing decisions are made. The process of get-ting dressed in the morning involves many decisionsabout garment selection and overall ensemble thermalproperties, and we expect that these decisions areinformed, in part, by our memory of what thermal envi-ronments were like outdoors yesterday, and possiblyalso what they are forecast to be like today. Evenoffice-workers who spend only a small fraction of theirwaking hours outdoors are exposed to the elements,and yesterday’s short-duration outdoor exposureexerts a major influence on their clothing decisions,over and above their expectations of the indoor cli-mates they are likely to encounter in the course of theirworking day.

When we more closely consider the effect exerted bytoday’s outdoor temperature on clothing decisions, wecome up against the question of just how can a dailymean temperature for a day that has not actually hap-pened yet exert an influence over our clothing selec-tions before we have even left home in the morning.Even the outdoor temperatures prevailing at the verysame time as our clothing decisions are being madeare, in many instances, yet to be experienced becauseour homes’ indoor climates are often divorced fromtheir outdoor environment. Part of the answer maysimply be that we intuitively expect today’s tempera-ture to be comparable to yesterday’s, but another pos-sibility may well be that we base our decisions onmeteorological forecasts, as disseminated through themass media which we routinely refer to at the time ofour clothing decisions. Breakfast TV and radio broad-

casts the world over disseminate the latest meteorolog-ical forecasts several times every hour, and morningnewspapers provide the same service in print.

To examine this hypothesis more closely, we ex-tracted Sydney’s daily maximum and minimum tem-perature forecasts issued in the very early morning ofeach day during the shopping mall study (12 August1996 through 31 January 1997). This involved scan-ning 6 mo of microfilm archives for Sydney’s maindaily newspaper (The Sydney Morning Herald). Theseforecasts would be similar if not identical to thoseissued to other media outlets such as radio and televi-sion, since their source was the same, namely the Aus-tralian Bureau of Meteorology. Therefore we are confi-dent that they are representative of the informationthat the subjects of our study were exposed to whenthey were making their clothing decisions in themorning.

Examining the ‘weather memory’ hypothesis first,47% of the day-to-day variance in average clo valuesin the shopping mall study was accounted for by yes-terday’s mean outdoor temperature, as indicated byFig. 6. The explained variance increased to 54% whentoday’s forecast mean outdoor temperature (i.e. meanof forecast max and min temperatures) was usedinstead of yesterday’s observations. Various combina-tions of independent models were run, but the highestscoring regression model combined both the ‘memory’and ‘weather forecast’ hypotheses together:

(2)

All terms in Eq. (2) were statistically significant (p <0.05), and 59% of the variance in day-to-day clo valuesis accounted for by the model described above (p <0.0001). It should be noted that the 2 ‘independentvariables’ in the model were themselves significantlycorrelated (R2 = 0.39), which renders the coefficients inthe model unstable. Nevertheless, this analysis lendsat least support to the ‘weather memory’ and ‘weatherforecast’ hypotheses of clothing decisions.

Fig. 6 represents another view on the ‘weather mem-ory’ hypothesis in relation to the shopping mall data. Inthis figure it can be noted that mean temperature onDay x had the strongest correlation with clothing insu-lation levels being worn in the shopping mall on Day x,with 52% of the variance being accounted for. But thecorrelation between clothing on Day x and tempera-ture on Day x−1 was also quite strong (R2 = 0.37), andeven Day x−2 exerted some influence (R2 = 0.20). It

today smean

clo value

yesterday sobserved mean

outdoor temperature

today sforecast

maximumtemperature

' '

'

. .

.

= −

−

1 15 0 0164

0 0178

279

Clim Res 24: 267–284, 2003

seems probable that part of the correlation from previ-ous days can be explained in terms of the autocorrela-tion occurring between successive days’ temperatures,but it also seems plausible that the subjects’ thermalmemory exerts an influence on clothing behaviour viathermal expectations.

4.6. Indoor clothing and energy consumption in thebuilt environment

In tropical, temperate and cold climate zones, thesingle largest impact exerted by an office building onits outdoor environment throughout its entire life cycleresults from the energy used creating a comfortableindoor climate (e.g. Roulet 2001). As noted by Holz etal. (1997) in a series of building energy simulation sen-sitivity analyses, clothing represents one of the mostsignificant comfort parameters influencing heatingand cooling loads. An attempt to quantify the signifi-cance, Newsham’s (1997) simulation study used a stan-dard dynamic energy simulation package (FENES-TRA) to ‘construct’ a virtual office space located inToronto, Canada. Heating, cooling and fan energyconsumption of the HVAC system were taken fromFENESTRA’s outputs. Thermal comfort conditionswere assessed in the simulated indoor climates usingthe conventional indices of predicted mean vote (PMV)and predicted percentage dissatisfied (PPD), calcu-lated for each time-step (Fanger 1970). The officebuilding simulation was run through a sample of daysfrom Toronto’s typical reference year (TRY) of meteo-rological data under a variety of operational scenarios.At one extreme was the ‘fixed-clothing’ scenario, inwhich the clo value of the hypothetical occupantswithin the space was prescribed, as per the usual com-fort standards such as ASHRAE Std 55 (ASHRAE1992). At the other extreme was the ‘variable-clothing’scenario, in which the occupants were permitted toadjust their clothing from a minimal socially accept-able level up to the maximum practical indoor level, asindicated by various published field studies of thermalcomfort.

Newsham’s analysis of total annual heating, coolingand fan energy was revealing (Newsham 1997). Treat-ing the heating, cooling and fan energy under the‘fixed-clothing’ scenario as 100%, it was reduced by 6,15 and 41% for the ‘very-limited-clothing-adjustment’,‘limited-clothing-adjustment’, and ‘variable-clothing-adjustment’ scenarios respectively. In the latter case,the thermostat deadband operating within the simu-lated building could be widened without compromis-ing comfort because more extreme temperatures couldbe accommodated by clothing adjustments. It is inter-esting to consider the implications of Newsham’s study

for the present results. Markee White’s (1986) studysuggests that an office environment with a formaldress code is the closest approximation of Newsham’s‘fixed clo’ scenario. The ‘casual Fridays’ (mufti days) inour Sydney office study would be akin to Newsham’s‘variable-clothing-adjustment’ scenario. So in theToronto context, heating, cooling and fan energy sav-ings accruing from the abandonment of a formal dresscode in office buildings could be of the order of 40%yr−1. Mapping these back-of-the-envelope calculationsto Sydney is not so simple, but if we accept that theenergy inputs to cooling are generally higher thanheating (due to differentials in thermodynamic effi-ciencies of the 2 processes), the 40% could well be asignificant underestimate in a warm climate zone. Ifnothing else, these speculations point towards somevery useful avenues for future research in which theseasonally varying clothing ensembles of the typeobserved during ‘casual Fridays’ in the present Sydneyoffice study are simulated within a software packagesuch as FENESTRA in a climate zone where HVACenergy is dominated by cooling loads. Such a studymight indicate the true costs in terms of energy andgreenhouse-gas emissions of ‘formal business attire’policies in the workplace.

The literature review at the start of this paper notedthe findings of the ASHRAE RP-884 adaptive thermalcomfort project (Figs. 1 & 2). There appeared to beroughly the same amount of indoor clothing insulationvariance explained by outdoor weather in the presentstudies (0.44 < R2 < 0.52 in Figs. 5 & 12) compared withearlier studies such as those depicted in Fig. 2 (R2 =0.49). This broad agreement occurred despite thewidely discrepant time scales of analysis in the 2 stud-ies. In the earlier adaptive model study, the insulation(clo) value in question represented the entire samplewithin a single building averaged for the duration ofthe study (typically a few weeks), while the outdoortemperature in that analysis was an average of thedaily maxima and minima across the same period.However, the outdoor temperature observations in thepresent shopping mall study were averaged across atypical daily sample size of about 45 subjects andlasted for 1 d only.

If outdoor conditions can account for about half thevariance in the level of clothing insulation worn insidebuildings when indoor and outdoor climates are virtu-ally disconnected, it becomes interesting to speculatewhat the percentage of explained clo variance mightbe in low-energy buildings, where indoor and outdoorclimates are more closely related. Free-running build-ings (Humphreys 1981) are those in which neitherheating nor air-conditioning are operating and thedegree of correlation between indoors and outdoorscan be quite high, depending on the thermal proper-

280

Morgan & de Dear: Weather, clothing and indoor climate

ties of the building envelope and ventilation rates. Thehypotheses and findings in this paper lead us to expectthat the linkage between indoor clo values and outdoorweather would be even stronger in free-running build-ings than we observed in air-conditioned settings.Clearly a corporate dress code has no place in suchbuildings, because it undermines this key mechanismof personal thermal adaptation. Indeed, this issue hasbeen flagged in the forthcoming revision to ASHRAEStd 55, which specifically precludes application of theadaptive model of thermal comfort in naturally venti-lated contexts where the building occupants are notfree to adjust their clothing to suit their own thermalcomfort preferences (ASHRAE 2002, de Dear & Brager2002).

The adaptive comfort standard (ACS) described indetail elsewhere (ASHRAE 2002, de Dear & Brager2002) prescribes comfortable and permissible indoortemperatures on the basis of concurrent outdoorweather conditions. The key independent variable inthe adaptive model underpinning ACS is the meanmonthly outdoor air temperature. In fact, the raw datathat went into the model used weather data integrationperiods ranging from a couple of days up to a couple ofweeks—in effect, the duration of each individualbuilding survey, of which there were 160 in the entireRP-884 database. The question of what is the optimaltime frame for the outdoor-temperature term in adap-tive models has not been thoroughly addressed to date.While the reference period of 1 mo was selected forimplementation in the ACS because of its widespreadavailability in published climatologies the world over,a more accurate time scale is probably shorter than this(Nicol et al. 1995). The findings in the present fieldstudies of clothing behaviour provide some guidanceas to what the optimal integration period might be. Inparticular, the correlation coefficients (or R2) for clovalues on Day x with outdoor temperatures on Day x,Day x–1, Day x–2, Day x –3 … Day x–n, as plotted inFig. 6, show a remarkably smooth exponential decaywith time, with a half-life of about 2 d. All coefficientsgoing back to Day x–7 were statistically significant atthe 0.05 level or better. This information provides arational basis for defining the most appropriate meanoutdoor temperature driver for adaptive models ofthermal comfort (Tmot). In our opinion, the ideal outdoortemperature function for adaptive comfort guidelinesis an exponentially weighted running mean spanningthe last 7 days. While today’s temperature, forecast orobserved, holds significant predictive power withregards to clothing insulation, it cannot be included insuch an adaptive algorithm, because the daily maxi-mum will not have occurred by the time it is actuallyneeded for calculation purposes. This 7 d integrationperiod is short enough to reflect the perceptual impacts

of recent weather dynamics, but also long enough tocapture ‘weather memory and persistence’ effects onhuman clothing behaviour.

The weighting coefficients for our proposedrunning-mean weekly outdoor temperature adaptivealgorithm (Tmot) were derived from the exponentialfunction plotted in Fig. 6. The actual calculations aredetailed in Table 3.

These weighting coefficients can be used to calcu-late the appropriate mean outdoor temperature (Tmot)for subsequent input to an adaptive indoor tempera-ture algorithm:

(3)

and then the adaptive algorithm for indoor comforttemperatures in a naturally ventilated or free runningbuilding (de Dear & Brager 2002) can be written as

comfort temperature (°C) = 0.31Tmot + 17.8 (4)

and the acceptable range of temperatures is defined asthe optimal comfort temperature (Eq. 4) ± 2.5°C for90% acceptability, or ±3.5°C for 80% acceptability(ASHRAE 2002).

Given that the adaptive model in Eq. (4) is relevantto naturally ventilated buildings, it may not be imme-diately apparent what purpose could be served by per-forming real-time calculations of temperature optima,or indeed acceptable temperature ranges, since thesebuildings, by definition, lack an HVAC responsemechanism. However, buildings with hybrid ventila-tion are very relevant to Eq. (4). The concept of hybridventilation is a simple one—the building operates innaturally ventilated mode whenever the outdooratmospheric environment is conducive to thermal com-

281

Day number x y = 0.69e−0.3688x Tmot weighting

Today 1 0.477 0.00Day x−1 2 0.330 0.34Day x–2 3 0.228 0.23Day x–3 4 0.158 0.16Day x–4 5 0.109 0.11Day x–5 6 0.075 0.08Day x–6 7 0.052 0.05Day x–7 8 0.036 0.03

1.00yx

x

=−

−

∑ 0 9891

7

.Day

Day

Table 3. Weighting coefficients for an adaptive algorithmbased on the exponential decay of outdoor weather influenceson indoor clothing insulation decisions. The variable y repre-sents the coefficient of determination (R2) for the relationshipbetween mean clo levels worn indoors and outdoor mean

daily temperatures (Fig. 6)

T T x T x T xT x T x T x T x

mot Day –1 Day –2 Day –3Day –4 Day –5 Day –6 Day –7

= + + ++ + +

0 34 0 23 0 160 11 0 08 0 05 0 03

. . .. . . .

Clim Res 24: 267–284, 2003

fort indoors, but then switches over to anactive HVAC system whenever outdoorconditions become too extreme to man-age passively (International EnergyAgency Annex 35 2001). Typically thecontrol of hybrid buildings is performedby a Building Management System (orEnergy Management System), and thesefacilities oftentimes have an automatedweather station on or near the building in question. Such a configuration canreadily compare continuously recordedindoor conditions with the correspondingrange of acceptable indoor temperatures(Eq. 4) based on the last 7 days’ outdoortemperatures (Eq. 3), and then switch thehybrid building over to active HVACmode whenever current conditions falloutside acceptable ranges.

An important caveat regarding thisproposed algorithm for Tmot is that it isonly relevant to the context in which itsunderlying data actually come from Sydney, Australia.There may well be different findings in other parts ofthe world, particularly where the day-to-day weatherand temperature fluctuations have a different ampli-tude and periodicity. The most comparable study to thepresent one was conducted by Nicol et al. (1995) inOxford, UK. In a 6 wk longitudinal thermal comfortresearch design, they estimated a half-life on the expo-nential decay relationship to be between 3 and 4 d, orabout twice as slow as that found in the current study.However, we are unsure whether this difference intime constant is due to differences in weather dynam-ics between Sydney and Oxford or the fact that thedependent variable in the Oxford study was thermalcomfort questionnaire vote and not clothing insulationvalue, as used in the current study.

Another caveat regarding our proposed algorithmfor Tmot is that the clothing data from which our 7 dweighting coefficients were derived (Eq. 3) may notcorrespond with ideal thermal comfort. We did not askour shopping mall subjects about their subjective ther-mal comfort states because direct contact with themwas precluded by our unobtrusive mode of observa-tion. In defence of the unobtrusive method we contendthat our subjects could not have been very uncomfort-able, because they were not motivated to change theirclothing to thermally more appropriate insulation lev-els. If they were seriously uncomfortable in the shop-ping mall the normal response would have been toleave. Nevertheless, the assumption of thermal com-fort states is less than ideal, so the Tmot equation (Eq. 3)developed out of our data could be improved withfuture research based on clothing behaviour in a set-

ting where thermal comfort can be assured moredirectly.

We also recognise that the utility of such an expo-nential function for the adaptive model is going to berestricted to applications in which real-time weatherobservations are available, typically from an AWS onthe same site as the building in question. In theabsence of real-time data there will be no practicalalternative to the climatological monthly norms (i.e.relying on the published mean monthly temperaturesfor input as Tmot to Eq. 4), as published by the localmeteorological service.

The Intergovernmental Panel on Climate Change(IPCC) reports that global average surface tempera-ture is projected to rise by 1.4 to 5.8°C by 2100, rela-tive to 1990 (Cubasch et al. 2001). In order to man-age this projected rise the United Nations ClimateChange Conference in Kyoto (1997) set in place theKyoto Protocol in which some developed nationsagreed to limit their greenhouse-gas emissions, rela-tive to the levels emitted in 1990. The implications ofthe present study for the issue of greenhouse-gasemissions are simple—that the ability to effectenergy conservation and reduce greenhouse-gasemissions attributable to the commercial buildingsector will be directly impacted by the degree offlexibility that building occupants have over theclothing they wear indoors. Newsham’s investigation(described earlier) into the relationship betweenclothing adjustments and energy savings in Torontobuildings (Newsham 1997) is just as relevant to theissue of greenhouse-gas emissions as it is to energyconsumption.

282

indoor climate

heating & air conditioning

building envelope

atmospheric environment• weather• climate• weather forecasts

thermo- regulation & thermal energy

balance

clothing decisions

recent experience & expectations

recent experience& expectations

heat and mass

externalenergy inputs

dress codes & policies

Fig. 13. Conceptual model of the interactions between weather/climate, the built environment and clothing

Morgan & de Dear: Weather, clothing and indoor climate

Finally, we hope that the research reported in thispaper might stimulate innovation within the clothingindustry. In particular, the paper provides a strong casefor the development of clothing garments that can fullycomply with formal office attire dress codes, yet pro-vide the wearer with enhanced adaptive opportunities.This might take the form of more variable clothing gar-ment combinations making up office-clothing ensem-bles. Alternatively, it may be possible to design insula-tion variability into individual garments, such asadjustable or even removable linings from jackets, orzippered vents concealed in the design and cut of jack-ets or dresses.

5. CONCLUSIONS

This paper has emphasised the impacts of outdoorweather, past (observed) and future (forecast), on thelevel of clothing insulation worn inside buildings.While the statistical association has been readilydemonstrated, the chain of causation has many linksand attenuating factors. Fig. 13 attempts to draw thiscomplexity together in the form of a conceptual modelof clothing, buildings, climate and energy. Startingwith weather and climate, the outdoor atmosphericenvironment impacts the energy balance of a givenbuilding, and that building’s indoor climate respondsaccordingly. The human occupants of the buildingmaintain their own energy balance with indoor cli-matic conditions, and the extent to which they rely onphysiologic (as opposed to behavioural or engineering)responses to maintain that energy balance determinesthe magnitude of their thermal discomfort and atten-dant dissatisfaction. A key behavioural mechanismthat attenuates thermal discomfort indoors is theadjustment of clothing insulation levels, but there areseveral factors, some of them psychological (cognitive,aesthetic, organisational, cultural, etc.) impinging onclothing decisions as well as the usual thermal vari-ables (temperature, humidity, etc.). One of these is theclothing policy of the occupants’ employer. Corporatedress codes, as found in the present office environmentstudy, all but extinguish clothing adaptive opportunity.The net result of such policies is to transfer responsibil-ity for comfort thermoregulation away from the indi-vidual and towards a building’s facilities manager.However, another factor impinging on clothing deci-sions is outdoor weather, either directly experienced orperhaps forecast by the local meteorological service.This linkage, if given free reign, offers potential toreduce our reliance on external energy inputs (air con-ditioning and heating) for the maintenance of thermalcomfort indoors. In so doing there is also likely to beattendant reductions in greenhouse-gas emissions

attributable to the commercial building sector. Energyconservation and greenhouse-gas emission reductionsare both premised on relaxation of the restrictionsapplying to the thermal adaptability of indoor clothingsuch as corporate dress codes.

Acknowledgements. The authors wish to thank Target storesand the Sitel Corporation for permission to conduct the studyat their Sydney premises. Meteorological data were suppliedby the Australian Bureau of Meteorology. Prof. Gail S. Bragerat the Center for Built Environment at the University of Cali-fornia, Berkeley, is thanked for her comments on variousaspects of this project.

LITERATURE CITED

ASHRAE (1992) ANSI/ASHRAE Standard 55—thermal envi-ronmental conditions for human occupancy. AmericanSociety of Heating, Refrigerating and Air-ConditioningEngineers, Atlanta

ASHRAE (1993) Physiological principles and thermal comfort.Handbook of fundamentals, Chap 8. American Society ofHeating, Refrigerating and Air-Conditioning Engineers,Atlanta

ASHRAE (1997) Physiological principles and thermal comfort,Chap 8. American Society of Heating, Refrigerating andAir-Conditioning Engineers, Atlanta

ASHRAE (2002) Public review draft of ANSI/ASHRAE Stan-dard 55R—thermal environmental conditions for humanoccupancy. American Society of Heating, Refrigeratingand Air-Conditioning Engineers, Atlanta

Auliciems A (1981) Towards a psycho-physiological model ofthermal perception. Int J Biometeorol 25:109−122

Auliciems A (1986) Air conditioning in Australia III: thermo-bile controls. Architect Sci Rev 33:43−48

Baker N, Standeven M (1996) Thermal comfort for free run-ning buildings. Energy Bldgs 23:175−182

Bureau of Meteorology (1991) Sydney: climatic survey. Aus-tralian Gov Pub Serv (AGPS), Canberra

Clark RP, Edholm OG (1985) Man and his thermal environ-ment. Edward Arnold, London

Cubasch U, Meehl GA, Boer GJ, Stouffer RJ, Dix M, Noda A,Senior CA, Raper S, Yap Y (2001) Projections of future cli-mate change. In: Houghton JT et al (eds) Climate change,2001, The scientific basis. Contribution of Working GroupI to the Third Assessment Report of the IntergovernmentalPanel on Climate Change. Cambridge University Press,Cambridge, p 267−289

Danielsson U (1993) Convection coefficients in clothing lay-ers. KTH doctoral thesis, University of Stockholm

de Dear RJ (1998) A global database of thermal comfort fieldexperiments. ASHRAE Trans 104:1141−1152

de Dear RJ, Brager G (1998) Developing an adaptive model ofthermal comfort and preference. ASHRAE Trans 104:145−167

de Dear RJ, Brager G (2001) The adaptive model of thermalcomfort and energy conservation in the built environment.Int J Biometeorol 45:100−108

de Dear RJ, Brager GS (2002) Thermal comfort in naturallyventilated buildings: revisions to ASHRAE Standard 55.Energy Bldgs 34:549−561

Fanger PO (1970) Thermal comfort. Danish Technical Press,Copenhagen

Fountain M, Brager G, de Dear RJ (1996) Expectations of

283

Clim Res 24: 267–284, 2003

indoor climate control. Energy Bldgs 24:179−182Gagge AP, Burton AC, Bazett HC (1941) A practical system of

units for the description of heat exchange of man with hisenvironment. Science 94:428−430

Gagge AP, Fobelets A, Berglund LG (1986) A standard pre-dictive index of human response to the thermal environ-ment. ASHRAE Trans 92:709−731

Goldman RF (1981) Evaluating the effects of clothing on thewearer. In: Cena K, Clark JA (eds) Bioengineering, ther-mal physiology and comfort. Elsevier, Amsterdam, p 41−55

Holz R, Hourigan A, Monkman SRP, Krarti M (1997) Effects ofstandard energy conserving measures on thermal comfort.Bldg Environ 32:31−42

Humphreys MA (1979) The influence of season and ambienttemperature on human clothing behaviour. In: Fanger PO(ed) Indoor climate. Danish Building Research Institute,Copenhagen, p 699−714

Humphreys MA (1981) The dependence of comfortable tem-peratures upon indoor and outdoor climates. In: Cena K,Clark JA (eds) Bioengineering, thermal physiology andcomfort. Elsevier, Amsterdam, p 229−250

International Energy Agency Annex 35 (2001) State of the artof hybrid ventilation, CD-ROM, Ver 2.0. InternationalEnergy Agency, Aalborg

Markee White NL (1986) Quantification of factors influencingthermal comfort in an office environment: implications forenergy conservation. Thesis, University of California atDavis

McCullough EA, Jones B (1984) A comprehensive databasefor estimating clothing insulation. Institute for Environ-mental Research, Manhattan, KS

McIntyre DA (1980) Indoor climate. Applied Science Publish-ers, London

Moser CA, Kalton G, (1971) Survey methods in social investi-gation. Heinemann Educational, London

Newsham G (1997) Clothing as a thermal moderator and theeffect on energy consumption. Energy Bldgs 26:283−291

Newsham G, Tiller DK (1997) A field study of office thermalcomfort using questionnaire software. ASHRAE Trans103:3–17

Nicol JF, Humphreys MA, Raja IA (1995) Developing indoortemperature standards for naturally ventilated buildings.CIBSE Conference. Chartered Institution of Building Ser-vices Engineers, London

Nishi Y, Gonzalez RR, Gagge AP (1975) Direct measurementof clothing heat transfer properties during sensible andinsensible heat exchange with the thermal environment.ASHRAE Trans 81:183−199

Olesen BW, Nielsen R (1983) Thermal insulation of clothingmeasured on a moveable manikin and on human subjects.Report, Technical University of Denmark, Lyngby

Parsons KC (1993) human thermal enviornments: the effectsof hot, moderate and cold environments on human health,comfort and performance. Taylor & Francis, London

Roulet CA (2001) Indoor environment quality in buildings andits impact on outdoor environment. Energy Bldgs 33:183−192

Rowe DM (1996) Some observations on thermal comfort in acommercial office building. Australian Institute of Refrig-eration, Heating and Air Conditioning International Con-ference, AIRAH, Hobart

Woodcock AH (1962) Moisture transfer in textile systems.Textile Res J 32:628−723

284

Editorial responsibility: Robert Davis (Editor),Charlottesville, Virginia, USA

Submitted: December 13, 2002; Accepted: May 21, 2003Proofs received from author(s): July 15, 2003