- Email: [email protected]
An adaptive thermal comfort model for the Tunisian context: a field study results
Energy and Buildings 37 (2005) 952–963 www.elsevier.com/locate/enbuild
An adaptive thermal comfort model for the Tunisian context: a field study results C. Bouden *, N. Ghrab Ecole Nationale d’Inge´nieurs de Tunis, B.P. 37-1002, Tunis Belve´de`re, Tunisia Received 10 September 2004; received in revised form 25 October 2004; accepted 10 December 2004
Abstract A thermal comfort field survey has been conducted in five towns from two climatic zones in Tunisia. Two hundred persons were concerned by this survey. They have been asked in their houses and working places at their normal living conditions once each month during 1 year. The selected sample of buildings has been chosen among free running buildings. Only few office buildings were equipped with heating systems but not with air conditioning systems. The survey results have been processed and correlations between sensations and globe temperatures have been found. The comfort temperatures have been calculated for each location and each month. They have been correlated with the mean outdoor air temperature and with the outdoors running mean temperature. Robust correlations have been found. The description of this survey and the methodology of data analysis are described in details in this paper. The main results are discussed and compared with those of similar surveys conducted in different geographic locations. # 2005 Elsevier B.V. All rights reserved. Keywords: Adaptive thermal comfort; Field survey; Free running buildings; Tunisian climate
1. Introduction In most of developed countries, the central heating systems are commonly used in winter. A control system is maintaining the indoor air temperature at a set value according to thermal comfort standards [1,18]. The use of thermal insulation, multiple glazing and controlled ventilation systems are necessary. In the North African countries, the climate is mild, but the needs to heating in winter and cooling in summer are certain, especially in coast areas where the humidity is high. The traditional buildings provided a relatively acceptable thermal comfort in winter and particularly in summer. These buildings are naturally ventilated and have a high thermal capacity. They are equipped with a courtyard; in those houses, adapted shadowing devices and appropriate orientations of the openings facilitate the control of the solar gains. During summer, when the outdoor air is cool, people * Corresponding author. Fax: +216 71 872 729. E-mail address: [email protected] (C. Bouden). 0378-7788/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.enbuild.2004.12.003
enjoy opening the windows and doors, the building is then cooled by natural ventilation and coolness is stored in the building thermal mass for the next day. This traditional way of construction is not anymore possible because of its cost and the change of people standard of life and behaviour; courtyard houses are replaced with flat or terrace houses. The development of ‘‘modern’’ construction was very fast and no care was paid to the thermal quality of the buildings. Most of the recent buildings are not equipped with thermal insulation and their tightness to air is very poor, this is due to the lack of appropriate standards. While the traditional buildings provide an acceptable thermal comfort, the recent buildings are very poor from the thermal point of view, and the thermal comfort is very low. During the last few years, the standard of life has risen, so people started to equip their houses and offices with heating systems in winter and with small air conditioning systems (mainly windows and split air conditioning systems) for the summer conditions. Therefore the energy consumption became very high because of the mediocre thermal quality of the buildings.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
In Tunisia, even if the buildings are heated and cooled, they are thermally controlled according to people sensation. The heating or air conditioning systems, if they exist, are put on or off manually when people feel uncomfortable. There is no set temperature for heating or cooling, and so the indoor temperatures are free running. Appropriate thermal standards for the North African countries, taking in consideration the particularities of the climate, the people culture and behaviour, and the method of occupation of the buildings, becomes necessary. Such standard is now under preparation. International thermal standards [1,18] are based on human energy balance obtained assuming steady state conditions. These standards are deduced from the experiments done by Fanger [13] in climatic chambers and use the predicted mean vote (PMV) equation to estimate the human mean response to the thermal environment from the knowledge of six thermal variables. Four among them are related to the ambience, they are the air temperature (Ta), the radiant temperature (Tr) or the globe temperature (Tg), the air velocity (Av) and the air relative humidity (RH) or vapour pressure ( pv). The two remaining parameters are related to the occupant’s adaptability to the local climate; these are the metabolic rate (M) and the clothing insulation (Icl). The steady state conditions are not reflecting the real life conditions particularly in North Africa where the indoor temperature is free running and the control is done manually by the occupant according to his feeling. People are used to adapt themselves to climate using clothing, posture, kind of activity, natural ventilation, etc. An assessment of the thermal comfort level as felt by the building’s occupants must be done under real occupation conditions. The objectives of this study are: 1. to investigate the occupants perception of the degree of comfort in dynamic conditions, 2. to examine the adaptive behaviour of the occupants on the usage of clothing, activity and available climatic controls in modifying the indoor thermal environment. 3. To develop an adaptive comfort model as an alternative to the fixed temperature set point that most of standards adopt. Such model fits better to the Tunisian context. 2. Needs for a thermal comfort survey The usual thermal comfort standards [1,18] have been established according to the results of experiments held in climatic chambers [13]. During these experiments, the person posture, his activity and his clothing are fixed, thus, the person sensation is taken under steady state conditions. In real conditions, these parameters are never fixed. Moreover, a person changes one or some of these parameters to adapt himself to his climatic conditions and to seasons. In Tunisia, three different climatic zones have been identified in the country. Each of these three zones has its own cultural heritage and traditional architecture reflecting its own particular climate. In addition, there are wide
953
seasonal changes in the weather between summer and winter. In summer, midday temperature can exceed 40 8C in some regions in the south. In winter, temperature can fall down to around 0 8C in the North Mountains. In the desert, at the country south, the freezing winter nights are also very common. The indoor design temperatures as described by international standards (ISO 7730 [18] and ASHRAE 551981, ASHRAE 55-1992 [1]) take no account of climatic variations and people adaptive behaviour. For any task and use of the building, there is a recommended temperature that is assumed to apply irrespective of climate and people culture, way of life and kind of clothing, though with some recognition of difference between summer and winter. Analysis of thermal comfort field studies have shown that indoor comfort temperature as felt by the occupants is function of mean outdoor temperature [15,2,22,24, 25,11,12,8]. This means that we can relate indoor comfort temperature to climate, region and seasons. For free running buildings and according to different surveys held under different climatic conditions, Humphreys [15] has found that the comfort temperature can be obtained from the mean outdoor temperature with Eq. (1) T c ¼ 0:534T o þ 11:9
(1)
Auliciems revised Humphreys equation by deleting some fields studies such those with children as the subjects, and adding more information from other studies not included by Humphreys [3,4]. These revisions increased the database to 53 separate field studies in various climatic zones covering more countries and more climates. After combining the data for naturally ventilated buildings and air-conditioned buildings, the analysis led to an equation involving the outdoors air temperature (To) and the indoor air temperature (Ti), this resulting equation is (Eq. (2)): T c ¼ 0:48T i þ 0:14T o þ 9:22
(2)
Auliciems has also proposed a single line for all buildings which covered the naturally ventilated buildings and airconditioned buildings [5]. This relation is given by Eq. (3) T c ¼ 0:31T o þ 17:6
(3)
Nicol has conducted several surveys under different climatic conditions. In a first survey in Pakistan [26], he has established a relation between comfort temperature and outdoor temperature given by Eq. (4) T c ¼ 0:38T o þ 17:0
(4)
In a second survey in Pakistan [27], he has found a second regression given by Eq. (5) T c ¼ 0:36T o þ 18:5
(5)
Those relations show clearly that the comfort temperature is related to the outdoor temperature and so to the climate. The difference between those relations confirms that there is no universal comfort temperature. Each community must have its own perception of the thermal comfort according to its climate, local culture and type of buildings. Such regression
954
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
does not exist for the North African climate and mode of occupation of the dwellings, this is why we considered to organize a survey through which we aim to learn more about people feeling of thermal comfort and methods of adaptations. A specific regression between comfort temperature and outdoor temperature is to be found. Standards based on such model would not only enhance thermal comfort but would also fit better to the local Tunisian and North African conditions (culture, behaviour, mode of occupation and nature of buildings) and would give energy saving.
3. Conduct of the survey and experimental method 3.1. Methodology
Table 1 Population sex repartition Women
Males
Total
Tunis
Houses Offices
12 10
8 10
20 20
Sfax
Houses Offices
13 10
7 10
20 20
Kef
Houses Offices
2 2
18 18
20 20
Gabe`s
Houses Offices
20 7
0 13
20 20
Gafsa
Houses Offices
14 7
6 13
20 20
Total
Houses Offices
61 36
39 64
100 100
97
103
200
Grand total
A cross-sectional investigation has been carried out monthly during a complete year in five different towns covering two climatic zones. These are: Kef, Tunis, Sfax, Gabe`s and Gafsa. Their approximate locations are shown in Fig. 1. In each town, 40 persons have been selected 20 among them have been interviewed in their houses and the 20 other persons have been interviewed in their working place. The total number of subjects was 200 persons spread in five cities. Each person has been visited 12 times, once in each month of the year. 3.2. The survey target population An equilibrium between males and females has been respected during the selection of the survey subjects;
Fig. 2. Population ages distribution.
Table 1 gives the subjects sex repartition for the different towns. The ages are evenly distributed from 20 years old to 70 years old. Fig. 2 gives the population age distribution for each region. The subjects have been selected by the researchers, which are in charge of the survey in each town. No more than two different persons were interviewed in the same building, hence, a larger number of different buildings has been covered. 3.3. Experimental method
Fig. 1. Location of the survey sites.
Four environmental variables were measured at the moment of the survey: the air temperature (Ta) in (8C), the globe temperature (Tg) in (8C), the relative humidity (RH) in (%) and the cooling time of a Kata thermometer (t) in (s). All sensors were positioned in the vicinity of the subject along the time during which the survey sheet is being filled. In most of the cases, the researcher filled the survey sheet. The questions in the survey sheet were written in French. Most of the subjects understood French. In some cases, the
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
researcher translates the questions. The survey sheet includes questions about: The subject sensation and preference about which he has to answer in a seven steps scale according to Bedford scale. His preference to be warmer or cooler. His skin moisture. His feeling and preference about the air velocity. His activity at the moment of the survey and 30 min before. The possibilities he has to control the ambience by opening a window, putting on a heater, a fan or a cooling device or controlling the solar radiation getting into the building. The subject has then to establish a list of the garment he is wearing at the moment of the survey by ticking in the appropriate box of a list including the most common clothes.
955
which the survey sheet has been filled; ADu is the subject body surface area calculated according to Dubois formula as function of the subject weight (W) and height (H) (see Eq. (8) from Mc Intyre [19])
A full description of the conduct of the survey and the results obtained from it is given in Refs. [6,7].
ADu ¼ 0:202W 0:425 H 0:725 ð1 met ¼ 58 W=m2 Þ (8) It seems reasonable to expect that the activity at the moment of the survey would be more influential than 30 min earlier. As far as it is known, no other investigators gave indication of appropriate weightings. Rowe [29] has conducted a similar survey in which he asked about the activity at the moment of the survey, 10 min before, 20 min before, 30 min before and 60 min before. He took the following ad hoc weighting factors: 50% for the 10 min preceding the moment of the survey, 25% for the 10 min before, 15% for the time corresponding to 30 min before the survey time and 10% for the time corresponding to one hour before the survey time. This method supposes that four values of the activity level are taken, which is, in our opinion, heavy and can lead to some uncertain responses. Rowe [29] has also added +10% to the metabolic rate in case the respondent has taken a snack or meal and +5% for a beverage.
3.4. Database
3.5. Globe temperatures or effective temperatures?
The responses from the questionnaires were entered in excel for analysis. A database has been built to filter the data according to any fixed criteria. During a preliminary data processing, we eliminated the records for which relevant informations are missing or inconsistent responses are encountered. The air velocity has been than calculated from the cooling time of the Kata thermometer, according to the appropriate equation for each thermometer.
Most of the papers encountered in the literature used the globe Temperature to establish the correlations and to find the comfort temperature. The idea of using a more aggregated parameter such as the effective temperature (ET*) or the standard effective temperature (SET) seems to be fascinating. The effective temperature (ET*) is defined as that temperature at 50% relative humidity, and mean radiant temperature equal to air temperature, that would produce the same thermal sensation as the actual environment. In other words, ET* normalises temperatures for humidity and radiation thereby facilitating comparison of various parameters with a single index. This parameter has been adopted by ASHRAE and appears in their comfort Standard 55 [1]. The ET* has been extended to incorporate different rates of activity and clothing insulation and became known as the standard effective temperature (SET). SET is described as a rationally derived index to indicate that it was obtained from an analysis of the physics of heat transfer. This distinguishes it from ET* which is considered as empirically derived. The effective temperature ET* has been calculated according to Eq. (9a) from Parsons [28]:
3.4.1. Calculation of the clothing insulation Individual clothing articles indicated in the survey responses were converted into their respective thermal insulation value (Ic) in units of Clo as tabulated by Mc Intyre [19] (1 Clo = 0.155 m2 C/W). The overall Clo value for each subject entire clothing ensemble has been calculated using the following equations from Mc Intyre [19] (Eqs. (6a) and (6b)): X I Clo;men ¼ 0:113 þ 0:727 Ic (6a) X I Clo;women ¼ 0:05 þ 0:77 Ic (6b) 3.4.2. Calculation of the metabolic rate The metabolic rate has been calculated from the activity level of each respondent at the moment in which the questionnaire sheet is being filled and his activity level 30 min before according to the following relation (Eq. (7)): ð0Þ
ð30Þ
M ¼ ADu 58ð0:7Iact þ 0:3Iact ð0Þ
ð30Þ
Þ
(7)
where Iact and Iact are the activity levels at the moment of the survey and respectively 30 min before the time during
ET ¼ 0:492T a þ 0:19ps þ 6:47 (9a) where pa is the vapour pressure (mbar) calculated as: 4030:18 pa ¼ RH exp 18:956 (9b) T a þ 235 where RH is the relative humidity and Ta is the ambient air temperature. The effective temperature has been calculated for the whole set of data. The comfort votes (S) have been correlated to ET* and to globe temperature. While the
956
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963 Table 2 Summary of the measured data distribution
Fig. 3. Regression of the effective temperature (ET*) to the globe temperature.
regression to ET* gave the correlation given by Eq. (10), the regression to the globe temperature gave a more robust correlation described by Eq. (11) ðR2 ¼ 0:6841Þ
(10)
S ¼ 0:1621T g 3:4482 ðR2 ¼ 0:7014Þ
(11)
S ¼ 0:259 ET 5:0962
The ET* has been correlated to the globe indoor temperature. The graph in Fig. 3 shows that those two parameters are highly correlated. We can also observe that the interval width of ET* is smaller than that of the globe temperature. This analysis has shown that the correlation to the globe temperature is more robust than that to ET*; then, we adopted to use the globe temperature for our investigation. This choice allows us to compare our results to many others from the literature. Bush [9] has shown that the use of ET* was particularly interesting when the population of the investigation is partly in naturally ventilated buildings and partly in air-conditioned buildings which is not our case. 3.6. Data summary Table 2 summarizes the distribution of the physical measured data by town.
4. Behavioural adaptation The clothing value, the metabolic rate, the posture and the air velocity indicate the behavioural adaptations. For the case of our study, the posture has not been taken in consideration. The only method that was found in the literature to measure the posture effect is very difficult to conduct in a so large and spread survey. Nicol [24] has conducted in Oxford Brookes University a 2-day survey in which he has shown that the change of posture with temperature causes a reduction of the body effective surface area of about 2% for each degree rise in temperature. Our survey focused on the evaluation of the clothing insulation and the metabolic rate. The air velocity was used as a means to measure the occupant control on the doors and windows opening.
Tair
Tglo
Humidity
ET*
Clo
Met
Gabe`s Max Min Average S.D.
29.5 15 23.5 3.9
29.2 15.2 24.1 4.0
77 34 56 8
25.9 15.6 21.2 2.7
1.22 0.07 0.55 0.28
2 0.7 1.23 0.28
Gafsa Max Min Average S.D.
35.5 11.9 21.6 5.9
28.1 12.4 21.9 5.8
70 18 52 9
23.3 14.3 19.6 3.4
1.43 0.16 0.62 0.28
2
Tunis Max Min Average S.D.
28.5 15.7 21.8 3.6
28.6 16.1 22.2 3.6
69 46 57 5
23.9 9.4 15.1 3.4
1.10 0.20 0,56 0.21
1.68
Sfax Max Min Average S.D.
31.2 14.5 22.7 4.7
31.2 15.2 22.8 4.7
79 43 64 6
28.4 16.0 21.1 3.2
1.54 0.12 0.53 0.25
1.98
Kef Max Min Average S.D.
39.2 4.8 20.45 9.16
39.5 4.5 20.5 9.2
98 4 63 15
31.4 10.1 19.5 5.5
1.80 0.14 0.73 0.37
2.8
1.19 0.22
1.12 0.12
1.17 0.21
1.24 0.34
4.1. Clothing value Clothing is an important behavioural adaptation to achieve thermal comfort at different temperatures. By plotting the value of the clothing thermal insulation worn by the survey subjects, we found that a correlation exist between the Clo value and both the outdoor temperature and the indoor globe temperature. Fig. 4a and b shows how the Clo value varies with the outdoor temperature respectively the indoor globe temperature. A very large scatter of the Clo value has been observed from 0.073 Clo in summer to 1.8 Clo in winter. A similar analysis has been led by de Dear [10], he has found the expression found by regression and given by Eq. (12) Clo ¼ 0:04T o þ 1:73
ðR2 ¼ 0:18Þ
(12)
Mui [21] in a similar study has found a very similar relation given by Eq. (13) Clo ¼ 0:04T o þ 1:76
ðR2 ¼ 0:21Þ
(13)
In our actual study, we found a regression (Eq. (14)) showing that the correlation between the clothing insulation and the outdoor temperature is very close to de Dear [10] and Mui [21] results, but the clothing overall insulation value in Clo was lower in our case. The constant term in our model is lower than those given in Eqs. (12) and (13). This correlation shows that the Tunisian population uses lighter clothes, and hence has a
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
957
4.2. Metabolic rate The plot of the metabolic rate versus temperature regression line shows that the regression line is nearly horizontal between 1.3 and 1.2 met independently of the value of the temperature. We can conclude that no significant correlation exists. This result sounds correct because the metabolic rate is mainly related to the nature of job or house activity rather than environment or temperature. One way of adaptation of the Tunisian population is to have only one working shift during the summer season, starting at 7:00 a.m. and ending at 2:00 p.m. During the afternoon hot hours, people are accustomed to take a nap, which reduces their metabolic rate. During our survey, we have never been at these hours to not disturb the people during their rest hours and so we cannot quantify this adaptive behaviour. 4.3. Air velocity
Fig. 4. (a) Correlation of clothing to outdoor temperature; (b) correlation of clothing to globe temperature.
higher physiological potential of adaptation to climate change than the populations investigated by de-Dear and Mui Clo ¼ 0:038T o þ 1:33 ðR2 ¼ 0:4919Þ
(14)
A second order polynomial regression for the same set of data gave Eq. (15). This latter seems to be more robust and visually fits better to the data scatter (see Fig. 4a) Clo ¼ 0:0017To2 0:1034T o þ 1:9
ðR2 ¼ 0:503Þ
(15)
Extrapolation of this equation to very high temperatures (40–50 8C), corresponding to desert conditions) shows that the clothing insulation increases. This result seems also realistic according to the behaviour of people living in the desert who wear heavy clothes during the hot summer days to protect themselves from the effects of hot air and solar radiation, and to favour the sweating. A similar regression has been done between the clothing and globe indoor temperature. Fig. 4b shows that the clothing varies with the globe indoor temperature. The linear regression gave the expression described by Eq. (16a), while the polynomial regression gave Eq. (16b) Clo ¼ 0:0352T g þ 1:3875
ðR2 ¼ 0:52Þ
Clo ¼ 0:0013Tg2 0:0909T g þ 1:95
ðR2 ¼ 0:57Þ
(16a) (16b)
We can conclude that a correlation exists between the clothing and the globe temperature. The correlation with indoor temperature is more robust than with outdoors temperature. The polynomial regression is 13% more robust than the linear regression.
The plot of the air velocity versus globe temperature has shown that this measurement is inconsistent. In fact, we can observe from the Kata thermometer measurement that the air velocity is in some cases very high since it can reach and even exceed 3 m/s, which is not possible inside a building. The average value was 0.4 m/s, which is also very high in our opinion for an average value. The standard deviation of the whole set of data is equal to 1.38. This result can be explained by the accuracy of the Kata thermometer and the care that the researchers have allocated to the measurement of the cooling time of the Kata thermometer. In fact, the Kata thermometers have been heated using hot water. During the cooling time, the thermometer has to be dry. We suspect that in some cases the researcher did not wipe the thermometer very well, which shortens the cooling time and gives us the impression that the air velocity is higher than it is in reality. We will consider that the air velocity is inconsistent in the case of our survey and we will not take it in consideration for our regressions.
5. Comfort votes and thermal acceptability Referring to previous research, the comfort temperature has been linearly related to outdoor temperature. The survey data have been used to generate the comfort temperature as felt by the subjects in their real living conditions, than the comfort temperatures have been correlated to the outdoor temperatures and to running mean temperatures. 5.1. Thermal acceptability The comfort votes of the population have been sorted in ascending order. The votes equal to (1), (0) and (1) are considered as those who take the operative temperature as
958
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
Fig. 5. (a) Histogram of the acceptable comfort votes distribution; (b) acceptable comfort votes as function of the globe temperature.
acceptable. For each value of the operative temperature, the total number of the votes and the percentage of the votes that are stating the acceptability have been calculated. Fig. 5a shows the acceptability as function of the operative temperature. We can notice that: 1. for temperatures between 16 and 26.5 8C, more than 80% are voting acceptable. This result shows that there is a high adaptation potential for a so wide temperature range, 2. the distribution is multi-modal. This result can be explained by the fact that for some values of low and some values of high temperatures, the total number of observations is very small, a unique vote stating acceptability will generate a high percentage for the corresponding value of the temperature. Those side percentages of acceptability made that a probit or logit regression is impossible. A polynomial regression has been tested. Two different equations have been found and their plots are shown in Fig. 5b. The function represented by an order 4 polynomial fits better the data scatter for the acceptable zone, unfortunately, this function deviates from the data for low and high temperatures, and this is to fit the marginal points caused by the small number of votes that have occurred in these temperatures. This function is also more robust, we can adopt it for lack of a better function as probit and we must pay attention to not use it for globe temperatures lower than 10 8C or higher than 35 8C.
5.2. Relation between PMV and sensation Using the measured parameters, the PMV has been calculated for each subject according to Fanger [13]. The values of the PMV have been compared to those of comfort votes (S) as described by the subjects. Fig. 6 gives the plot of the PMV as function of the comfort votes. This figure shows that for each value of the sensation (comfort votes), the scatter of the PMV is very large. The regression line does not correspond to the line ‘‘ y = x’’. For the value of S = 3, the value of the PMV from the regression is also equal to 3. For S = 0, the PMV (from the regression) is equal to 0.87 and for S = +3, the PMV (from regression) is equal to 1.28 which is very far from the comfort vote. A polynomial regression seems to be more robust and visually fits better
Fig. 6. Calculated PMV as function of comfort votes.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
than the linear regression, but also exhibits a scatter from the line ‘‘ y = x’’. The conclusion that can be inferred from this analysis is that the PMV is a measure of thermal comfort which has been deduced from steady state conditions measurement held in climatic chambers. This measure does not exactly correspond to the real sensation of people particularly in free running buildings in which people are accustomed to adapt themselves to climate. A more realistic model taking in consideration the people potential of adaptation is necessary for free running buildings and particularly in the Tunisian climate and context.
6. Development of the adaptive model Referring to previous research, the comfort temperature has been linearly related to outdoors temperature. The survey data have been used to generate the comfort temperature as felt by the subjects in their real living conditions, then the comfort temperatures have been correlated to the outdoors temperatures and to the running mean temperatures. 6.1. Comfort temperature correlation The comfort temperature can be defined as the indoors operative temperature for which people can vote comfortable on the Bedford scale (neutral on the ASHRAE scale). For each city and each month, the scatter of the comfort votes (sensation) has been plotted versus the globe temperature. It was observed that some of the points exhibit a bias from the scatter of points. A statistical method called the standard residual method has been used to detect the inconsistent points. Those point which are having a standard residual higher than 2 are automatically eliminated. For each point, the standard residual is calculated as: ei R¼ sˆ u where ei is the residual calculated as: ei ¼ Y i Yˆ i sffiffiffiffiffiffiffiffiffiffiffiffi ðei Þ2 sˆ u ¼ N2 Yi is the point ordinate and Yˆ i is the ordinate of the corresponding point on the regression line. According to this method 63 data among 1537 have been removed, which correspond to 4% of the whole set of data. The analysis will be held using 1474 records. A linear regression according to the least-squares method has been established between the comfort votes (sensations) and the indoors operative temperature for each town and each month. The comfort temperature is found from the intersection between the regression lines and the neutral
959
sensation line. The determination coefficient R2 is calculated to determine the robustness of the correlation. It outcomes that for some months, the regression line never crosses the neutrality line. For the hot summer months, the indoor globe temperature are very often high and people sensations are higher than ‘‘S = 0’’ (S = 0 corresponds to neutrality). The comfort temperature cannot be inferred from such set of data. In the mid-seasons, the temperatures are very close to the comfort temperature, most of the votes are taking the value ‘‘S = 0’’ for different values of the temperature, the regression line is nearly horizontal in this case. In such situation, the exact value for the comfort temperature cannot also be inferred and the determination coefficient is very low, which means that the correlation is very weak. Many methods have been adopted to get round this problem: 1. The data can be gathered in a seasonal basis and a thermal comfort temperature can be inferred for each season. 2. The data can be gathered in an annual basis, and an annual thermal comfort temperature can be calculated for each location. This will hide the adaptive behaviour of the occupants along the year. 3. The comfort temperature can be inferred from the Griffiths method [14] which takes an assumed regression slope of 0.33/8C for the relation between comfort votes and temperature. This method has been adopted and used later by Humphreys and Nicol [17]. 4. The comfort temperature can be inferred from the method proposed by Brager and de Dear [8] who have revised the Griffiths method. They made an extensive inspection of the ASHRAE database of thermal comfort field studies and using this data, they suggested an appropriate slope of 0.5 for the relationship between the comfort votes and temperature. This new value of the slope has been adopted by Mc Cartney and Nicol [20] to develop the adaptive control algorithm when they analysed the data for the smart controls and thermal comfort (SCATs) project. Hence, according to these last two methods, the comfort votes (subjects sensations) can be correlated to the operative temperature as: S = a. Tg + b, where a = 0.33 according to Griffiths or a = 0.5 according to de Dear and Brager. The comfort temperature will be defined as the temperature corresponding to neutrality, and can be calculated for each vote. A monthly mean comfort temperature can than inferred. 6.2. Correlations to outdoor running mean temperatures In 1978, Humphreys [15] has shown that the outdoor monthly mean outdoor temperature can be used as an appropriate outdoor temperature index. Later studies from Nicol and Humphreys [24,16] have shown that an exponentially weighted mean outdoor temperature can give
960
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
more accurate results. This index has been called the ‘‘running mean temperature’’ it calculates a weighted-mean from the last day and the previous days with decreasing weighing factors. The expression of the running mean temperature is given by Eq. (17) TRMðnÞ ¼ C TRMðn1Þ þ ð1 CÞ TMOðn1Þ
(17)
(n)
where TRM is the running mean temperature for the actual day; TRM(n 1) is the running mean temperature 1 day before; TMO(n 1) is the mean outdoor temperature 1 day before; C is a constant, its value is between 0 and 1 and defines the quickening response of the running mean to changes in the outdoors temperature. A value close to 0 gives a running mean temperature close to yesterday’ mean outdoor temperature and a value close to 1 gives a running mean temperature equal to the average value of the last days. This running mean temperature is similar to half-life decay calculations in nuclear physics. Humphreys showed that the half-life of a particular running mean temperature is approximately equal to 0.69/ (1 C). For a value of C = 0.31, the half-life is 1 day, and for a value of C = 0.9, the half-life is 1 week. In [20], Mc Kartney and Nicol have shown that the value of 0.8 gives the best results for the SCATs project.
7. Data analysis and results 7.1. Seasonal regressions The data were grouped on a season basis. For each city, four seasons were considered. Despite the similar globe temperatures in spring and autumn, both seasons were considered separately. We observed that warm temperatures were more accepted as comfortable in autumn than in spring. This observation can be explained by two facts: In autumn, people are accustomed to high temperatures after having experienced 3 months of summer. In autumn, the people clothing value is lower than in spring during which, people are still wearing heavy clothes. The classification of data in four seasons made that in each season, the range of temperature is larger than it was in a monthly basis, and the sample is much larger: up to 120 observations per city. The regressions were more robust, and the values of the comfort temperatures were found for each season without extrapolation of the regression line. In some cities, some months were missing (the case of Sfax in April for example), this made that some seasonal sets of data are having smaller populations in such case. The set of data is not representative for the whole season. 7.2. Yearly regressions The data have been grouped on an annual basis for each town. Each set of data corresponding to a city is composed of
Table 3 Coefficients of the regression line ‘‘S = aTg + b’’ calculated on a yearly basis City
a
b
R2
Kef Tunis Sfax Gabe`s Gafsa
0.1661 0.1568 0.1559 0.1105 0.1675
3.62 3.18 3.27 2.31 3.60
0.84 0.50 0.57 0.29 0.74
up to 480 data representing the whole year. This made the regression more robust with determination coefficients higher than 0.5 in most of the cases (except Gabe`s). The globe temperature covers a wide range from 8.3 to 34.3 8C. Table 3 shows the results of the regressions ‘‘S = aTg + b’’. We can observe that Kef and Gafsa data are very similar: both cities are far from the coast and they have comparable values for a, b and R2. A regression line calculated for the data of both cities together gave the following expression: ‘‘S = 0.1669Tg + 0.3889; R2 = 0.7984’’. For the three remaining cities which are on the coasts, we can observe that the regression coefficients are decreasing from north to south (Tunis: 0.157; Sfax: 0.156; Gabe`s: 0.1105). This result can be explained by the fact that people from the North of Tunisia exhibit a higher sensibility to temperature change than people leaving in the south of the country. The observation of the set of data has shown that for temperatures ranging from 15 to 30 8C, the minimum vote was ‘‘slightly cold’’ corresponding to ‘‘S = 1’’ in Tunis, while it was ‘‘Cold’’ corresponding to ‘‘S = 2’’ in Sfax and ‘‘Very cold’’ corresponding to ‘‘S = 3’’ in Gabe`s. For temperatures ranging around 30 8C, some people in Tunis have voted ‘‘very hot’’, ‘‘S = 3’’; in Sfax and Gabe`s they have voted ‘‘hot’’, which corresponds to ‘‘S = 2’’. This means that people leaving in the south of Tunisia resist more to high temperatures than people leaving in the North of the country. 7.3. Correlation between Griffith’s comfort temperature and outdoors temperature Because the number of subjects is very small in each town and because of the narrow range of temperature in each month and town, the comfort temperature estimated from the regression (S = f(Tg)) could not be relied upon. Comfort temperatures were therefore calculated using the Griffiths method [14]. This method is very well adapted for small sets of data. For each subject, the comfort temperature according to Griffiths method has been determined. For each town and each month, the average comfort temperature has been calculated. In other words, for each town, each set of monthly data corresponding to a survey is averaged in one data point. The 12 points corresponding to the 12 months have been correlated to the monthly mean outdoor temperature. The results found from these correlations are given in Table 4.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963 Table 4 Coefficients for the regression: ‘‘Tc-Griffiths = aTo-Avg + b’’
7.4. Correlation to running mean temperature
City
a
b
R2
Kef Tunis Sfax Gabe`s Gafsa All Tunisia
0.7472 0.4132 0.44 0.4314 0.4593 0.5179
7.75 13.43 13.41 14.12 12.72 10.35
0.956 0.785 0.889 0.814 0.954 0.96
For the all five cities, the results of the linear regression are given by Eq. (18) T c-Griffiths ¼ 0:518T o-Avg þ 10:35
(18)
The regression results (Table 4) show that four towns among five are very close to each others. The Kef correlation is different, this latter shows that Kef people are having a higher adaptation capacity than people from the other regions in that sense that they accept a lower comfort temperature when outdoor temperature is very low and a higher comfort temperature in summer. All the correlations for the five times show that the comfort temperature is strongly correlated to outdoor temperature. The determination coefficients show the robustness of the correlations. The correlations obtained according to de Dear and Brager method confirm those results. de Dear and Brager correlations show that the comfort temperatures are highly correlated to the mean outdoor temperature. The determination coefficients show that these last correlations are 3% more robust than those obtained according to Griffiths method. The correlation slopes are all higher than those obtained by Griffiths method. The results found from de Dear and Brager are given in Table 5. For the all five cities, the results of the linear regression are given by the expression of Eq. (19) T c-Brager ¼ 0:68T o-Avg þ 6:88
961
(19)
For the survey range of outdoor temperatures (8–35 8C), the correlations found for the five cities together are comparable to the one found by Humphreys [15] and described by Eq. (1) and to those found by Nicol in Pakistan’s surveys [26,27] and given by Eqs. (4) and (5). Among all correlations, the one done according to de Dear and Brager is having the steepest slope. It shows a high adaptive potential of the population vis-a`-vis the outdoor temperature. Table 5 Coefficients for the regression: ‘‘Tc-Brager = aTo-Avg + b’’ City
a
b
R2
Kef Tunis Sfax Gabe`s Gafsa All Tunisia
0.9977 0.5089 0.5597 0.5248 0.6044 0.6758
3.06 11.93 11.40 12.46 9.95 6.88
0.97 0.88 0.95 0.92 0.97 0.99
Referring to Mc Kartney and Nicol [20], the running mean temperature has been calculated with a coefficient C = 0.8. The Griffiths comfort temperatures have been correlated to the daily running mean temperatures. The resulting correlation is given by the expression in Eq. (20) T c-Griffiths ¼ 0:532 TRM þ 11:56
ðR2 ¼ 0:6262Þ
(20)
This result is very close to the correlation between comfort temperature and mean outdoor temperature, but has a slightly higher determination coefficient. Compared to the results of Mc Kartney and Nicol [20], this equation shows that the Tunisian population is having a higher adaptation capacity than the average European population that was involved in the SCATs project. This difference can be explained by the fact that the SCATs project has been held in air-conditioned buildings while the Tunisian survey covered only naturally ventilated buildings. The correlation is very close to the one found by Humphreys in [15], and does not justify the use of the running mean temperature because it was very close to the one correlated to the outdoor temperature. The determination coefficient found for the running mean temperature regression was 35% lower than that found for the mean outdoor temperature regression. 7.5. Prediction of the comfort temperatures Both equations found from regression according to Griffiths method and de Dear and Brager method are having good determination coefficients respectively 0.96 and 0.99. It is tempting to say that those two expressions should be used to predict the indoor comfort temperature from the monthly mean outdoor temperature particularly because they were very close to the expression given by Humphreys [15]. The expression given by Auliciem and de Dear is found from data covering naturally ventilated buildings and airconditioned buildings. Differences of up to 6 8C were found between the predictions made using the correlation found from our survey’s results and the correlation given by de Dear and Auliciem in Ref. [11]. This difference between both correlations is important especially for low values of the temperature. This difference can be explained by the fact that Auliciem’s correlation takes in consideration both kinds of buildings. From the other side, Nicol found in two different surveys that he has organized in Pakistan [23,24], two correlations very close to the one found by Auliciem despite of the fact that in Pakistan, the buildings are in most of the cases naturally ventilated. The correlation found from de Dear and Brager method seems more robust than the one found according to Griffiths method, but for low outdoor temperatures, the comfort temperature found from Griffiths sounds more plausible. In fact, for an outdoor temperature of 8 8C, the de Dear and
962
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
Brager correlation predicts a comfort temperature of 12.3 8C which is in our opinion very low unless people wear very heavy clothes which is very restricting for their activity.
8. Conclusion Building standards have been based on fixed comfort temperatures found from tests held in climatic chambers. Those standards assume that the indoor temperature is fixed to a set value and controlled by heating and air conditioning systems. In Tunisia and all North African countries, the heating and air conditioning systems, in case they exist, are not used continuously. Thus, the indoor temperature is fluctuating. The thermal sensation of the building occupants is the only controller of the ventilation, the heating or the cooling of the building. Unlike the conventional thermal regulations which are based on energy consumption, the special feature of the future Tunisian thermal regulation is related to the fact that it must ensure a minimum level of thermal comfort when the building is free running without any heating or cooling system. This study has shown that the Tunisian population has a large potential of adaptation to climate and seasons. Correlation between clothing and temperature relates such potential of adaptation. The comfort temperature can be correlated to the monthly mean outdoor temperature. Such concept can be used to design comfortable buildings. The comparison between the comfort temperature and the maximum and minimum outdoors temperatures can help designer to judge whether passive-heating and cooling techniques are appropriate for the climate under consideration. The relationship between the indoor comfort temperature and the range of outdoors temperatures shows whether for example night ventilation can be or not a way to keep the building cool in summer and help the designer to select the appropriate thermal capacity for the building. Our study suggests that in Tunisia, the thermal comfort temperature can be calculated from one of the expressions: T c-Griffiths ¼ 0:518T o-Avg þ 10:35; T c-Brager ¼ 0:680T o-Avg þ 6:88 where Tc-Griffiths (8C) is the comfort temperature calculated using the Griffiths method, Tc-Brager (8C) is the comfort temperature found according to de Dear and Brager Method, To-Avg (8C) is the monthly mean outdoors temperature. The correlations found in the case of our study are in good agreement with those found in the literature particularly the one found by Humphreys. For buildings equipped with heating and air conditioning systems, a variable indoors temperature has to be taken according to the comfort temperature calculated from the
above equation. This is in agreement with the common way people used to run their heating or cooling systems. Such method of running will allow an important energy saving compared with the existing standards. Our study is a first step towards a thermal comfort standard adapted for the Tunisian conditions. In order to confirm these results, further surveys have to be carried out to validate the results on a larger population.
Acknowledgments The survey has been financed by the ‘‘Agence Nationale pour les Energies Renouvelables’’ in Tunisia and was supported by the British Council. Dr. Fergus Nicol and Rev. M.A. Humphreys have contributed by their directives and advise to the selection of the survey instruments, the preparation of the questionnaires and the data processing. They are gratefully acknowledged for all their support.
References [1] ASHRAE: Standard 55, Thermal environmental conditions for human occupancy, American Society of Heating Refrigeration and Air Conditioning Engineers, Atlanta, GA, USA, 1992. [2] A. Auliciem, R. de Dear, Air conditioning in Australia. I. Human factors, Architectural Science Review 29 (1978) 67–75. [3] A. Auliciems, Towards a psychophisiological model of thermal perception, International Journal of Biometeorology 13 (1981) 147–162. [4] A. Auliciems, Psycho-physiological criteria for global zones of building design, in: Proceedings of the Ninth International Society of Biometeorology Conference, Stuttgart, Hohenheim, 1983. [5] A. Auliciems, R. de Dear, Air conditioning in Australia: human thermal factors, Architectural Science Review 29 (1986) 67–75. [6] C. Bouden, N. Ghrab-Morcos, J.F. Nicol, M. Humphreys, A thermal comfort survey in Tunisia, in: Proceedings of the Second European Conference on Energy Performance and Indoor Climate in Buildings, vol. 2, Lyon, France, 1998, , pp. 491–496. [7] C. Bouden, N. Ghrab-Morcos, Le confort thermique dans les baˆ timents tunisiens: Re´ sultats d’une enqueˆ te, rapport interne publie´ a` l’ENIT, TN, 1999. [8] G.S. Brager, R.J. de Dear, Developing an adaptive model of thermal comfort and adaptation, in: ASHRAE Technical Data Bulletin, vol. 14, No. 1, ASHRAE, Atlanta, USA, 1998. [9] J.F. Bush, A tale of two populations: thermal comfort in air conditioned and naturally ventilated offices in Thailand, Energy and Buildings 18 (1992) 235–249. [10] R. de Dear, B.G. Schiller, D. Cooper, Developing an adaptive model of thermal comfort and preference, ASHRAE Transactions 104 (1) (1988), SF-98-7-3: 4106, RP 884. [11] R. de Dear, A. Auliciem, A field study of occupant comfort and office environments in hot-humid climate, Final report ASHRAE RP-702, 1993. [12] R.J. de Dear, G.S. Brager, Towards an adaptive model of thermal comfort and preference, ASHRAE Transactions 104 (1) (1998). [13] P.O. Fanger, Thermal Comfort, Danish Technical Press, Copenhagen, 1970. [14] I. Griffiths, Thermal comfort studies in buildings with passive solar features: field studies report to commission of the European Community, ENS 35909, UK, 1990.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963 [15] M.A. Humphreys, Outdoor temperatures and comfort indoors, Building Research and Practice 6 (2) (1978) 92–105. [16] M.A. Humphreys, J.F. Nicol, J.F. Nicol, et al. An Adaptive Temperature Guidelines for UK Office Temperatures. Standard for Thermal Comfort: Indoor Air Temperature Standards for the 21st Century, E & F Spon, London, UK, 1995. [17] M.A. Humphreys, J.F. Nicol, Outdoor temperature and indoor thermal comfort: rising the precision on the relationship for the 1998 ASHRAE database of fields studies, ASHRAE Transactions 206 (2) (2000) 485– 492. [18] ISO, International Standard 7730, Moderate Thermal Environments: Determination of PMV and PPD Indices and Specification of the Conditions for Thermal Comfort, International Organisation for Standardisation, Geneva, 1994. [19] D.A. Mc Intyre, Indoor Climate, Applied Science Publisher Ltd., London, 1980. [20] K.J. Mc Cartney, J.F. Nicol, Developing an adaptive control algorithm for Europe, Energy and Buildings 34 (2002) 623–635. [21] K.W.H. Mui, W.T.D. Chan, Adaptive comfort temperature model of air conditioned buildings in Hong Kong, Building and Environment 38 (2003) 837–852. [22] J.F. Nicol, A. Auliciem, A survey of thermal comfort in Pakistan toward new indoor temperature standards, Final Report to the Over-
[23]
[24]
[25] [26]
[27]
[28]
[29]
963
seas Development, Administration Published by Oxford Brookes University, School of Architecture, UK, 1994. J.F. Nicol, I. Raja, A. Alauddin, Thermal comfort in Pakistan II-toward new indoor temperature standards, Report to the Overseas Development Administration, Oxford Brookes University, UK, 1996. J.F. Nicol, I.A. Raja, Thermal comfort and posture. Exploratory studies in the nature of adaptive thermal comfort, Oxford Brookes University, School of Architecture, 1996. ISBN: 1-89940403-1. J.F. Nicol, S. Roaf, Pioneering new indoor temperature standards: the Pakistan project, Energy in Buildings 23 (3) (1996) 169–174. J.F. Nicol, Thermal comfort and temperature standards in Pakistan, in: J.F. Nicol, M. Humphreys, O. Sykes, S. Roaf (Eds.), Standards for Thermal Comfort–Indoor Air Temperature Standards for the 21st Century, E & F Spon, London, 1995, 149–157. J.F. Nicol, I.A. Raja, A. Alauddin, N.G. Jamy, Climatic variations on comfortable temperature: the Pakistan projects, Energy and Buildings 30 (1999) 261–279. K.C. Parsons, Human Thermal Environments: The Effects of Hot, Moderate and Cold Environments on Human Health, Comfort and Performance, Taylor and Francis, 1993. D.M. Rowe, Activity rates and thermal comfort of office occupants in Sydney, Journal of Thermal Biology 26 (2001) 415–418.
An adaptive thermal comfort model for the Tunisian context: a field study results C. Bouden *, N. Ghrab Ecole Nationale d’Inge´nieurs de Tunis, B.P. 37-1002, Tunis Belve´de`re, Tunisia Received 10 September 2004; received in revised form 25 October 2004; accepted 10 December 2004
Abstract A thermal comfort field survey has been conducted in five towns from two climatic zones in Tunisia. Two hundred persons were concerned by this survey. They have been asked in their houses and working places at their normal living conditions once each month during 1 year. The selected sample of buildings has been chosen among free running buildings. Only few office buildings were equipped with heating systems but not with air conditioning systems. The survey results have been processed and correlations between sensations and globe temperatures have been found. The comfort temperatures have been calculated for each location and each month. They have been correlated with the mean outdoor air temperature and with the outdoors running mean temperature. Robust correlations have been found. The description of this survey and the methodology of data analysis are described in details in this paper. The main results are discussed and compared with those of similar surveys conducted in different geographic locations. # 2005 Elsevier B.V. All rights reserved. Keywords: Adaptive thermal comfort; Field survey; Free running buildings; Tunisian climate
1. Introduction In most of developed countries, the central heating systems are commonly used in winter. A control system is maintaining the indoor air temperature at a set value according to thermal comfort standards [1,18]. The use of thermal insulation, multiple glazing and controlled ventilation systems are necessary. In the North African countries, the climate is mild, but the needs to heating in winter and cooling in summer are certain, especially in coast areas where the humidity is high. The traditional buildings provided a relatively acceptable thermal comfort in winter and particularly in summer. These buildings are naturally ventilated and have a high thermal capacity. They are equipped with a courtyard; in those houses, adapted shadowing devices and appropriate orientations of the openings facilitate the control of the solar gains. During summer, when the outdoor air is cool, people * Corresponding author. Fax: +216 71 872 729. E-mail address: [email protected] (C. Bouden). 0378-7788/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.enbuild.2004.12.003
enjoy opening the windows and doors, the building is then cooled by natural ventilation and coolness is stored in the building thermal mass for the next day. This traditional way of construction is not anymore possible because of its cost and the change of people standard of life and behaviour; courtyard houses are replaced with flat or terrace houses. The development of ‘‘modern’’ construction was very fast and no care was paid to the thermal quality of the buildings. Most of the recent buildings are not equipped with thermal insulation and their tightness to air is very poor, this is due to the lack of appropriate standards. While the traditional buildings provide an acceptable thermal comfort, the recent buildings are very poor from the thermal point of view, and the thermal comfort is very low. During the last few years, the standard of life has risen, so people started to equip their houses and offices with heating systems in winter and with small air conditioning systems (mainly windows and split air conditioning systems) for the summer conditions. Therefore the energy consumption became very high because of the mediocre thermal quality of the buildings.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
In Tunisia, even if the buildings are heated and cooled, they are thermally controlled according to people sensation. The heating or air conditioning systems, if they exist, are put on or off manually when people feel uncomfortable. There is no set temperature for heating or cooling, and so the indoor temperatures are free running. Appropriate thermal standards for the North African countries, taking in consideration the particularities of the climate, the people culture and behaviour, and the method of occupation of the buildings, becomes necessary. Such standard is now under preparation. International thermal standards [1,18] are based on human energy balance obtained assuming steady state conditions. These standards are deduced from the experiments done by Fanger [13] in climatic chambers and use the predicted mean vote (PMV) equation to estimate the human mean response to the thermal environment from the knowledge of six thermal variables. Four among them are related to the ambience, they are the air temperature (Ta), the radiant temperature (Tr) or the globe temperature (Tg), the air velocity (Av) and the air relative humidity (RH) or vapour pressure ( pv). The two remaining parameters are related to the occupant’s adaptability to the local climate; these are the metabolic rate (M) and the clothing insulation (Icl). The steady state conditions are not reflecting the real life conditions particularly in North Africa where the indoor temperature is free running and the control is done manually by the occupant according to his feeling. People are used to adapt themselves to climate using clothing, posture, kind of activity, natural ventilation, etc. An assessment of the thermal comfort level as felt by the building’s occupants must be done under real occupation conditions. The objectives of this study are: 1. to investigate the occupants perception of the degree of comfort in dynamic conditions, 2. to examine the adaptive behaviour of the occupants on the usage of clothing, activity and available climatic controls in modifying the indoor thermal environment. 3. To develop an adaptive comfort model as an alternative to the fixed temperature set point that most of standards adopt. Such model fits better to the Tunisian context. 2. Needs for a thermal comfort survey The usual thermal comfort standards [1,18] have been established according to the results of experiments held in climatic chambers [13]. During these experiments, the person posture, his activity and his clothing are fixed, thus, the person sensation is taken under steady state conditions. In real conditions, these parameters are never fixed. Moreover, a person changes one or some of these parameters to adapt himself to his climatic conditions and to seasons. In Tunisia, three different climatic zones have been identified in the country. Each of these three zones has its own cultural heritage and traditional architecture reflecting its own particular climate. In addition, there are wide
953
seasonal changes in the weather between summer and winter. In summer, midday temperature can exceed 40 8C in some regions in the south. In winter, temperature can fall down to around 0 8C in the North Mountains. In the desert, at the country south, the freezing winter nights are also very common. The indoor design temperatures as described by international standards (ISO 7730 [18] and ASHRAE 551981, ASHRAE 55-1992 [1]) take no account of climatic variations and people adaptive behaviour. For any task and use of the building, there is a recommended temperature that is assumed to apply irrespective of climate and people culture, way of life and kind of clothing, though with some recognition of difference between summer and winter. Analysis of thermal comfort field studies have shown that indoor comfort temperature as felt by the occupants is function of mean outdoor temperature [15,2,22,24, 25,11,12,8]. This means that we can relate indoor comfort temperature to climate, region and seasons. For free running buildings and according to different surveys held under different climatic conditions, Humphreys [15] has found that the comfort temperature can be obtained from the mean outdoor temperature with Eq. (1) T c ¼ 0:534T o þ 11:9
(1)
Auliciems revised Humphreys equation by deleting some fields studies such those with children as the subjects, and adding more information from other studies not included by Humphreys [3,4]. These revisions increased the database to 53 separate field studies in various climatic zones covering more countries and more climates. After combining the data for naturally ventilated buildings and air-conditioned buildings, the analysis led to an equation involving the outdoors air temperature (To) and the indoor air temperature (Ti), this resulting equation is (Eq. (2)): T c ¼ 0:48T i þ 0:14T o þ 9:22
(2)
Auliciems has also proposed a single line for all buildings which covered the naturally ventilated buildings and airconditioned buildings [5]. This relation is given by Eq. (3) T c ¼ 0:31T o þ 17:6
(3)
Nicol has conducted several surveys under different climatic conditions. In a first survey in Pakistan [26], he has established a relation between comfort temperature and outdoor temperature given by Eq. (4) T c ¼ 0:38T o þ 17:0
(4)
In a second survey in Pakistan [27], he has found a second regression given by Eq. (5) T c ¼ 0:36T o þ 18:5
(5)
Those relations show clearly that the comfort temperature is related to the outdoor temperature and so to the climate. The difference between those relations confirms that there is no universal comfort temperature. Each community must have its own perception of the thermal comfort according to its climate, local culture and type of buildings. Such regression
954
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
does not exist for the North African climate and mode of occupation of the dwellings, this is why we considered to organize a survey through which we aim to learn more about people feeling of thermal comfort and methods of adaptations. A specific regression between comfort temperature and outdoor temperature is to be found. Standards based on such model would not only enhance thermal comfort but would also fit better to the local Tunisian and North African conditions (culture, behaviour, mode of occupation and nature of buildings) and would give energy saving.
3. Conduct of the survey and experimental method 3.1. Methodology
Table 1 Population sex repartition Women
Males
Total
Tunis
Houses Offices
12 10
8 10
20 20
Sfax
Houses Offices
13 10
7 10
20 20
Kef
Houses Offices
2 2
18 18
20 20
Gabe`s
Houses Offices
20 7
0 13
20 20
Gafsa
Houses Offices
14 7
6 13
20 20
Total
Houses Offices
61 36
39 64
100 100
97
103
200
Grand total
A cross-sectional investigation has been carried out monthly during a complete year in five different towns covering two climatic zones. These are: Kef, Tunis, Sfax, Gabe`s and Gafsa. Their approximate locations are shown in Fig. 1. In each town, 40 persons have been selected 20 among them have been interviewed in their houses and the 20 other persons have been interviewed in their working place. The total number of subjects was 200 persons spread in five cities. Each person has been visited 12 times, once in each month of the year. 3.2. The survey target population An equilibrium between males and females has been respected during the selection of the survey subjects;
Fig. 2. Population ages distribution.
Table 1 gives the subjects sex repartition for the different towns. The ages are evenly distributed from 20 years old to 70 years old. Fig. 2 gives the population age distribution for each region. The subjects have been selected by the researchers, which are in charge of the survey in each town. No more than two different persons were interviewed in the same building, hence, a larger number of different buildings has been covered. 3.3. Experimental method
Fig. 1. Location of the survey sites.
Four environmental variables were measured at the moment of the survey: the air temperature (Ta) in (8C), the globe temperature (Tg) in (8C), the relative humidity (RH) in (%) and the cooling time of a Kata thermometer (t) in (s). All sensors were positioned in the vicinity of the subject along the time during which the survey sheet is being filled. In most of the cases, the researcher filled the survey sheet. The questions in the survey sheet were written in French. Most of the subjects understood French. In some cases, the
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
researcher translates the questions. The survey sheet includes questions about: The subject sensation and preference about which he has to answer in a seven steps scale according to Bedford scale. His preference to be warmer or cooler. His skin moisture. His feeling and preference about the air velocity. His activity at the moment of the survey and 30 min before. The possibilities he has to control the ambience by opening a window, putting on a heater, a fan or a cooling device or controlling the solar radiation getting into the building. The subject has then to establish a list of the garment he is wearing at the moment of the survey by ticking in the appropriate box of a list including the most common clothes.
955
which the survey sheet has been filled; ADu is the subject body surface area calculated according to Dubois formula as function of the subject weight (W) and height (H) (see Eq. (8) from Mc Intyre [19])
A full description of the conduct of the survey and the results obtained from it is given in Refs. [6,7].
ADu ¼ 0:202W 0:425 H 0:725 ð1 met ¼ 58 W=m2 Þ (8) It seems reasonable to expect that the activity at the moment of the survey would be more influential than 30 min earlier. As far as it is known, no other investigators gave indication of appropriate weightings. Rowe [29] has conducted a similar survey in which he asked about the activity at the moment of the survey, 10 min before, 20 min before, 30 min before and 60 min before. He took the following ad hoc weighting factors: 50% for the 10 min preceding the moment of the survey, 25% for the 10 min before, 15% for the time corresponding to 30 min before the survey time and 10% for the time corresponding to one hour before the survey time. This method supposes that four values of the activity level are taken, which is, in our opinion, heavy and can lead to some uncertain responses. Rowe [29] has also added +10% to the metabolic rate in case the respondent has taken a snack or meal and +5% for a beverage.
3.4. Database
3.5. Globe temperatures or effective temperatures?
The responses from the questionnaires were entered in excel for analysis. A database has been built to filter the data according to any fixed criteria. During a preliminary data processing, we eliminated the records for which relevant informations are missing or inconsistent responses are encountered. The air velocity has been than calculated from the cooling time of the Kata thermometer, according to the appropriate equation for each thermometer.
Most of the papers encountered in the literature used the globe Temperature to establish the correlations and to find the comfort temperature. The idea of using a more aggregated parameter such as the effective temperature (ET*) or the standard effective temperature (SET) seems to be fascinating. The effective temperature (ET*) is defined as that temperature at 50% relative humidity, and mean radiant temperature equal to air temperature, that would produce the same thermal sensation as the actual environment. In other words, ET* normalises temperatures for humidity and radiation thereby facilitating comparison of various parameters with a single index. This parameter has been adopted by ASHRAE and appears in their comfort Standard 55 [1]. The ET* has been extended to incorporate different rates of activity and clothing insulation and became known as the standard effective temperature (SET). SET is described as a rationally derived index to indicate that it was obtained from an analysis of the physics of heat transfer. This distinguishes it from ET* which is considered as empirically derived. The effective temperature ET* has been calculated according to Eq. (9a) from Parsons [28]:
3.4.1. Calculation of the clothing insulation Individual clothing articles indicated in the survey responses were converted into their respective thermal insulation value (Ic) in units of Clo as tabulated by Mc Intyre [19] (1 Clo = 0.155 m2 C/W). The overall Clo value for each subject entire clothing ensemble has been calculated using the following equations from Mc Intyre [19] (Eqs. (6a) and (6b)): X I Clo;men ¼ 0:113 þ 0:727 Ic (6a) X I Clo;women ¼ 0:05 þ 0:77 Ic (6b) 3.4.2. Calculation of the metabolic rate The metabolic rate has been calculated from the activity level of each respondent at the moment in which the questionnaire sheet is being filled and his activity level 30 min before according to the following relation (Eq. (7)): ð0Þ
ð30Þ
M ¼ ADu 58ð0:7Iact þ 0:3Iact ð0Þ
ð30Þ
Þ
(7)
where Iact and Iact are the activity levels at the moment of the survey and respectively 30 min before the time during
ET ¼ 0:492T a þ 0:19ps þ 6:47 (9a) where pa is the vapour pressure (mbar) calculated as: 4030:18 pa ¼ RH exp 18:956 (9b) T a þ 235 where RH is the relative humidity and Ta is the ambient air temperature. The effective temperature has been calculated for the whole set of data. The comfort votes (S) have been correlated to ET* and to globe temperature. While the
956
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963 Table 2 Summary of the measured data distribution
Fig. 3. Regression of the effective temperature (ET*) to the globe temperature.
regression to ET* gave the correlation given by Eq. (10), the regression to the globe temperature gave a more robust correlation described by Eq. (11) ðR2 ¼ 0:6841Þ
(10)
S ¼ 0:1621T g 3:4482 ðR2 ¼ 0:7014Þ
(11)
S ¼ 0:259 ET 5:0962
The ET* has been correlated to the globe indoor temperature. The graph in Fig. 3 shows that those two parameters are highly correlated. We can also observe that the interval width of ET* is smaller than that of the globe temperature. This analysis has shown that the correlation to the globe temperature is more robust than that to ET*; then, we adopted to use the globe temperature for our investigation. This choice allows us to compare our results to many others from the literature. Bush [9] has shown that the use of ET* was particularly interesting when the population of the investigation is partly in naturally ventilated buildings and partly in air-conditioned buildings which is not our case. 3.6. Data summary Table 2 summarizes the distribution of the physical measured data by town.
4. Behavioural adaptation The clothing value, the metabolic rate, the posture and the air velocity indicate the behavioural adaptations. For the case of our study, the posture has not been taken in consideration. The only method that was found in the literature to measure the posture effect is very difficult to conduct in a so large and spread survey. Nicol [24] has conducted in Oxford Brookes University a 2-day survey in which he has shown that the change of posture with temperature causes a reduction of the body effective surface area of about 2% for each degree rise in temperature. Our survey focused on the evaluation of the clothing insulation and the metabolic rate. The air velocity was used as a means to measure the occupant control on the doors and windows opening.
Tair
Tglo
Humidity
ET*
Clo
Met
Gabe`s Max Min Average S.D.
29.5 15 23.5 3.9
29.2 15.2 24.1 4.0
77 34 56 8
25.9 15.6 21.2 2.7
1.22 0.07 0.55 0.28
2 0.7 1.23 0.28
Gafsa Max Min Average S.D.
35.5 11.9 21.6 5.9
28.1 12.4 21.9 5.8
70 18 52 9
23.3 14.3 19.6 3.4
1.43 0.16 0.62 0.28
2
Tunis Max Min Average S.D.
28.5 15.7 21.8 3.6
28.6 16.1 22.2 3.6
69 46 57 5
23.9 9.4 15.1 3.4
1.10 0.20 0,56 0.21
1.68
Sfax Max Min Average S.D.
31.2 14.5 22.7 4.7
31.2 15.2 22.8 4.7
79 43 64 6
28.4 16.0 21.1 3.2
1.54 0.12 0.53 0.25
1.98
Kef Max Min Average S.D.
39.2 4.8 20.45 9.16
39.5 4.5 20.5 9.2
98 4 63 15
31.4 10.1 19.5 5.5
1.80 0.14 0.73 0.37
2.8
1.19 0.22
1.12 0.12
1.17 0.21
1.24 0.34
4.1. Clothing value Clothing is an important behavioural adaptation to achieve thermal comfort at different temperatures. By plotting the value of the clothing thermal insulation worn by the survey subjects, we found that a correlation exist between the Clo value and both the outdoor temperature and the indoor globe temperature. Fig. 4a and b shows how the Clo value varies with the outdoor temperature respectively the indoor globe temperature. A very large scatter of the Clo value has been observed from 0.073 Clo in summer to 1.8 Clo in winter. A similar analysis has been led by de Dear [10], he has found the expression found by regression and given by Eq. (12) Clo ¼ 0:04T o þ 1:73
ðR2 ¼ 0:18Þ
(12)
Mui [21] in a similar study has found a very similar relation given by Eq. (13) Clo ¼ 0:04T o þ 1:76
ðR2 ¼ 0:21Þ
(13)
In our actual study, we found a regression (Eq. (14)) showing that the correlation between the clothing insulation and the outdoor temperature is very close to de Dear [10] and Mui [21] results, but the clothing overall insulation value in Clo was lower in our case. The constant term in our model is lower than those given in Eqs. (12) and (13). This correlation shows that the Tunisian population uses lighter clothes, and hence has a
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
957
4.2. Metabolic rate The plot of the metabolic rate versus temperature regression line shows that the regression line is nearly horizontal between 1.3 and 1.2 met independently of the value of the temperature. We can conclude that no significant correlation exists. This result sounds correct because the metabolic rate is mainly related to the nature of job or house activity rather than environment or temperature. One way of adaptation of the Tunisian population is to have only one working shift during the summer season, starting at 7:00 a.m. and ending at 2:00 p.m. During the afternoon hot hours, people are accustomed to take a nap, which reduces their metabolic rate. During our survey, we have never been at these hours to not disturb the people during their rest hours and so we cannot quantify this adaptive behaviour. 4.3. Air velocity
Fig. 4. (a) Correlation of clothing to outdoor temperature; (b) correlation of clothing to globe temperature.
higher physiological potential of adaptation to climate change than the populations investigated by de-Dear and Mui Clo ¼ 0:038T o þ 1:33 ðR2 ¼ 0:4919Þ
(14)
A second order polynomial regression for the same set of data gave Eq. (15). This latter seems to be more robust and visually fits better to the data scatter (see Fig. 4a) Clo ¼ 0:0017To2 0:1034T o þ 1:9
ðR2 ¼ 0:503Þ
(15)
Extrapolation of this equation to very high temperatures (40–50 8C), corresponding to desert conditions) shows that the clothing insulation increases. This result seems also realistic according to the behaviour of people living in the desert who wear heavy clothes during the hot summer days to protect themselves from the effects of hot air and solar radiation, and to favour the sweating. A similar regression has been done between the clothing and globe indoor temperature. Fig. 4b shows that the clothing varies with the globe indoor temperature. The linear regression gave the expression described by Eq. (16a), while the polynomial regression gave Eq. (16b) Clo ¼ 0:0352T g þ 1:3875
ðR2 ¼ 0:52Þ
Clo ¼ 0:0013Tg2 0:0909T g þ 1:95
ðR2 ¼ 0:57Þ
(16a) (16b)
We can conclude that a correlation exists between the clothing and the globe temperature. The correlation with indoor temperature is more robust than with outdoors temperature. The polynomial regression is 13% more robust than the linear regression.
The plot of the air velocity versus globe temperature has shown that this measurement is inconsistent. In fact, we can observe from the Kata thermometer measurement that the air velocity is in some cases very high since it can reach and even exceed 3 m/s, which is not possible inside a building. The average value was 0.4 m/s, which is also very high in our opinion for an average value. The standard deviation of the whole set of data is equal to 1.38. This result can be explained by the accuracy of the Kata thermometer and the care that the researchers have allocated to the measurement of the cooling time of the Kata thermometer. In fact, the Kata thermometers have been heated using hot water. During the cooling time, the thermometer has to be dry. We suspect that in some cases the researcher did not wipe the thermometer very well, which shortens the cooling time and gives us the impression that the air velocity is higher than it is in reality. We will consider that the air velocity is inconsistent in the case of our survey and we will not take it in consideration for our regressions.
5. Comfort votes and thermal acceptability Referring to previous research, the comfort temperature has been linearly related to outdoor temperature. The survey data have been used to generate the comfort temperature as felt by the subjects in their real living conditions, than the comfort temperatures have been correlated to the outdoor temperatures and to running mean temperatures. 5.1. Thermal acceptability The comfort votes of the population have been sorted in ascending order. The votes equal to (1), (0) and (1) are considered as those who take the operative temperature as
958
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
Fig. 5. (a) Histogram of the acceptable comfort votes distribution; (b) acceptable comfort votes as function of the globe temperature.
acceptable. For each value of the operative temperature, the total number of the votes and the percentage of the votes that are stating the acceptability have been calculated. Fig. 5a shows the acceptability as function of the operative temperature. We can notice that: 1. for temperatures between 16 and 26.5 8C, more than 80% are voting acceptable. This result shows that there is a high adaptation potential for a so wide temperature range, 2. the distribution is multi-modal. This result can be explained by the fact that for some values of low and some values of high temperatures, the total number of observations is very small, a unique vote stating acceptability will generate a high percentage for the corresponding value of the temperature. Those side percentages of acceptability made that a probit or logit regression is impossible. A polynomial regression has been tested. Two different equations have been found and their plots are shown in Fig. 5b. The function represented by an order 4 polynomial fits better the data scatter for the acceptable zone, unfortunately, this function deviates from the data for low and high temperatures, and this is to fit the marginal points caused by the small number of votes that have occurred in these temperatures. This function is also more robust, we can adopt it for lack of a better function as probit and we must pay attention to not use it for globe temperatures lower than 10 8C or higher than 35 8C.
5.2. Relation between PMV and sensation Using the measured parameters, the PMV has been calculated for each subject according to Fanger [13]. The values of the PMV have been compared to those of comfort votes (S) as described by the subjects. Fig. 6 gives the plot of the PMV as function of the comfort votes. This figure shows that for each value of the sensation (comfort votes), the scatter of the PMV is very large. The regression line does not correspond to the line ‘‘ y = x’’. For the value of S = 3, the value of the PMV from the regression is also equal to 3. For S = 0, the PMV (from the regression) is equal to 0.87 and for S = +3, the PMV (from regression) is equal to 1.28 which is very far from the comfort vote. A polynomial regression seems to be more robust and visually fits better
Fig. 6. Calculated PMV as function of comfort votes.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
than the linear regression, but also exhibits a scatter from the line ‘‘ y = x’’. The conclusion that can be inferred from this analysis is that the PMV is a measure of thermal comfort which has been deduced from steady state conditions measurement held in climatic chambers. This measure does not exactly correspond to the real sensation of people particularly in free running buildings in which people are accustomed to adapt themselves to climate. A more realistic model taking in consideration the people potential of adaptation is necessary for free running buildings and particularly in the Tunisian climate and context.
6. Development of the adaptive model Referring to previous research, the comfort temperature has been linearly related to outdoors temperature. The survey data have been used to generate the comfort temperature as felt by the subjects in their real living conditions, then the comfort temperatures have been correlated to the outdoors temperatures and to the running mean temperatures. 6.1. Comfort temperature correlation The comfort temperature can be defined as the indoors operative temperature for which people can vote comfortable on the Bedford scale (neutral on the ASHRAE scale). For each city and each month, the scatter of the comfort votes (sensation) has been plotted versus the globe temperature. It was observed that some of the points exhibit a bias from the scatter of points. A statistical method called the standard residual method has been used to detect the inconsistent points. Those point which are having a standard residual higher than 2 are automatically eliminated. For each point, the standard residual is calculated as: ei R¼ sˆ u where ei is the residual calculated as: ei ¼ Y i Yˆ i sffiffiffiffiffiffiffiffiffiffiffiffi ðei Þ2 sˆ u ¼ N2 Yi is the point ordinate and Yˆ i is the ordinate of the corresponding point on the regression line. According to this method 63 data among 1537 have been removed, which correspond to 4% of the whole set of data. The analysis will be held using 1474 records. A linear regression according to the least-squares method has been established between the comfort votes (sensations) and the indoors operative temperature for each town and each month. The comfort temperature is found from the intersection between the regression lines and the neutral
959
sensation line. The determination coefficient R2 is calculated to determine the robustness of the correlation. It outcomes that for some months, the regression line never crosses the neutrality line. For the hot summer months, the indoor globe temperature are very often high and people sensations are higher than ‘‘S = 0’’ (S = 0 corresponds to neutrality). The comfort temperature cannot be inferred from such set of data. In the mid-seasons, the temperatures are very close to the comfort temperature, most of the votes are taking the value ‘‘S = 0’’ for different values of the temperature, the regression line is nearly horizontal in this case. In such situation, the exact value for the comfort temperature cannot also be inferred and the determination coefficient is very low, which means that the correlation is very weak. Many methods have been adopted to get round this problem: 1. The data can be gathered in a seasonal basis and a thermal comfort temperature can be inferred for each season. 2. The data can be gathered in an annual basis, and an annual thermal comfort temperature can be calculated for each location. This will hide the adaptive behaviour of the occupants along the year. 3. The comfort temperature can be inferred from the Griffiths method [14] which takes an assumed regression slope of 0.33/8C for the relation between comfort votes and temperature. This method has been adopted and used later by Humphreys and Nicol [17]. 4. The comfort temperature can be inferred from the method proposed by Brager and de Dear [8] who have revised the Griffiths method. They made an extensive inspection of the ASHRAE database of thermal comfort field studies and using this data, they suggested an appropriate slope of 0.5 for the relationship between the comfort votes and temperature. This new value of the slope has been adopted by Mc Cartney and Nicol [20] to develop the adaptive control algorithm when they analysed the data for the smart controls and thermal comfort (SCATs) project. Hence, according to these last two methods, the comfort votes (subjects sensations) can be correlated to the operative temperature as: S = a. Tg + b, where a = 0.33 according to Griffiths or a = 0.5 according to de Dear and Brager. The comfort temperature will be defined as the temperature corresponding to neutrality, and can be calculated for each vote. A monthly mean comfort temperature can than inferred. 6.2. Correlations to outdoor running mean temperatures In 1978, Humphreys [15] has shown that the outdoor monthly mean outdoor temperature can be used as an appropriate outdoor temperature index. Later studies from Nicol and Humphreys [24,16] have shown that an exponentially weighted mean outdoor temperature can give
960
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
more accurate results. This index has been called the ‘‘running mean temperature’’ it calculates a weighted-mean from the last day and the previous days with decreasing weighing factors. The expression of the running mean temperature is given by Eq. (17) TRMðnÞ ¼ C TRMðn1Þ þ ð1 CÞ TMOðn1Þ
(17)
(n)
where TRM is the running mean temperature for the actual day; TRM(n 1) is the running mean temperature 1 day before; TMO(n 1) is the mean outdoor temperature 1 day before; C is a constant, its value is between 0 and 1 and defines the quickening response of the running mean to changes in the outdoors temperature. A value close to 0 gives a running mean temperature close to yesterday’ mean outdoor temperature and a value close to 1 gives a running mean temperature equal to the average value of the last days. This running mean temperature is similar to half-life decay calculations in nuclear physics. Humphreys showed that the half-life of a particular running mean temperature is approximately equal to 0.69/ (1 C). For a value of C = 0.31, the half-life is 1 day, and for a value of C = 0.9, the half-life is 1 week. In [20], Mc Kartney and Nicol have shown that the value of 0.8 gives the best results for the SCATs project.
7. Data analysis and results 7.1. Seasonal regressions The data were grouped on a season basis. For each city, four seasons were considered. Despite the similar globe temperatures in spring and autumn, both seasons were considered separately. We observed that warm temperatures were more accepted as comfortable in autumn than in spring. This observation can be explained by two facts: In autumn, people are accustomed to high temperatures after having experienced 3 months of summer. In autumn, the people clothing value is lower than in spring during which, people are still wearing heavy clothes. The classification of data in four seasons made that in each season, the range of temperature is larger than it was in a monthly basis, and the sample is much larger: up to 120 observations per city. The regressions were more robust, and the values of the comfort temperatures were found for each season without extrapolation of the regression line. In some cities, some months were missing (the case of Sfax in April for example), this made that some seasonal sets of data are having smaller populations in such case. The set of data is not representative for the whole season. 7.2. Yearly regressions The data have been grouped on an annual basis for each town. Each set of data corresponding to a city is composed of
Table 3 Coefficients of the regression line ‘‘S = aTg + b’’ calculated on a yearly basis City
a
b
R2
Kef Tunis Sfax Gabe`s Gafsa
0.1661 0.1568 0.1559 0.1105 0.1675
3.62 3.18 3.27 2.31 3.60
0.84 0.50 0.57 0.29 0.74
up to 480 data representing the whole year. This made the regression more robust with determination coefficients higher than 0.5 in most of the cases (except Gabe`s). The globe temperature covers a wide range from 8.3 to 34.3 8C. Table 3 shows the results of the regressions ‘‘S = aTg + b’’. We can observe that Kef and Gafsa data are very similar: both cities are far from the coast and they have comparable values for a, b and R2. A regression line calculated for the data of both cities together gave the following expression: ‘‘S = 0.1669Tg + 0.3889; R2 = 0.7984’’. For the three remaining cities which are on the coasts, we can observe that the regression coefficients are decreasing from north to south (Tunis: 0.157; Sfax: 0.156; Gabe`s: 0.1105). This result can be explained by the fact that people from the North of Tunisia exhibit a higher sensibility to temperature change than people leaving in the south of the country. The observation of the set of data has shown that for temperatures ranging from 15 to 30 8C, the minimum vote was ‘‘slightly cold’’ corresponding to ‘‘S = 1’’ in Tunis, while it was ‘‘Cold’’ corresponding to ‘‘S = 2’’ in Sfax and ‘‘Very cold’’ corresponding to ‘‘S = 3’’ in Gabe`s. For temperatures ranging around 30 8C, some people in Tunis have voted ‘‘very hot’’, ‘‘S = 3’’; in Sfax and Gabe`s they have voted ‘‘hot’’, which corresponds to ‘‘S = 2’’. This means that people leaving in the south of Tunisia resist more to high temperatures than people leaving in the North of the country. 7.3. Correlation between Griffith’s comfort temperature and outdoors temperature Because the number of subjects is very small in each town and because of the narrow range of temperature in each month and town, the comfort temperature estimated from the regression (S = f(Tg)) could not be relied upon. Comfort temperatures were therefore calculated using the Griffiths method [14]. This method is very well adapted for small sets of data. For each subject, the comfort temperature according to Griffiths method has been determined. For each town and each month, the average comfort temperature has been calculated. In other words, for each town, each set of monthly data corresponding to a survey is averaged in one data point. The 12 points corresponding to the 12 months have been correlated to the monthly mean outdoor temperature. The results found from these correlations are given in Table 4.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963 Table 4 Coefficients for the regression: ‘‘Tc-Griffiths = aTo-Avg + b’’
7.4. Correlation to running mean temperature
City
a
b
R2
Kef Tunis Sfax Gabe`s Gafsa All Tunisia
0.7472 0.4132 0.44 0.4314 0.4593 0.5179
7.75 13.43 13.41 14.12 12.72 10.35
0.956 0.785 0.889 0.814 0.954 0.96
For the all five cities, the results of the linear regression are given by Eq. (18) T c-Griffiths ¼ 0:518T o-Avg þ 10:35
(18)
The regression results (Table 4) show that four towns among five are very close to each others. The Kef correlation is different, this latter shows that Kef people are having a higher adaptation capacity than people from the other regions in that sense that they accept a lower comfort temperature when outdoor temperature is very low and a higher comfort temperature in summer. All the correlations for the five times show that the comfort temperature is strongly correlated to outdoor temperature. The determination coefficients show the robustness of the correlations. The correlations obtained according to de Dear and Brager method confirm those results. de Dear and Brager correlations show that the comfort temperatures are highly correlated to the mean outdoor temperature. The determination coefficients show that these last correlations are 3% more robust than those obtained according to Griffiths method. The correlation slopes are all higher than those obtained by Griffiths method. The results found from de Dear and Brager are given in Table 5. For the all five cities, the results of the linear regression are given by the expression of Eq. (19) T c-Brager ¼ 0:68T o-Avg þ 6:88
961
(19)
For the survey range of outdoor temperatures (8–35 8C), the correlations found for the five cities together are comparable to the one found by Humphreys [15] and described by Eq. (1) and to those found by Nicol in Pakistan’s surveys [26,27] and given by Eqs. (4) and (5). Among all correlations, the one done according to de Dear and Brager is having the steepest slope. It shows a high adaptive potential of the population vis-a`-vis the outdoor temperature. Table 5 Coefficients for the regression: ‘‘Tc-Brager = aTo-Avg + b’’ City
a
b
R2
Kef Tunis Sfax Gabe`s Gafsa All Tunisia
0.9977 0.5089 0.5597 0.5248 0.6044 0.6758
3.06 11.93 11.40 12.46 9.95 6.88
0.97 0.88 0.95 0.92 0.97 0.99
Referring to Mc Kartney and Nicol [20], the running mean temperature has been calculated with a coefficient C = 0.8. The Griffiths comfort temperatures have been correlated to the daily running mean temperatures. The resulting correlation is given by the expression in Eq. (20) T c-Griffiths ¼ 0:532 TRM þ 11:56
ðR2 ¼ 0:6262Þ
(20)
This result is very close to the correlation between comfort temperature and mean outdoor temperature, but has a slightly higher determination coefficient. Compared to the results of Mc Kartney and Nicol [20], this equation shows that the Tunisian population is having a higher adaptation capacity than the average European population that was involved in the SCATs project. This difference can be explained by the fact that the SCATs project has been held in air-conditioned buildings while the Tunisian survey covered only naturally ventilated buildings. The correlation is very close to the one found by Humphreys in [15], and does not justify the use of the running mean temperature because it was very close to the one correlated to the outdoor temperature. The determination coefficient found for the running mean temperature regression was 35% lower than that found for the mean outdoor temperature regression. 7.5. Prediction of the comfort temperatures Both equations found from regression according to Griffiths method and de Dear and Brager method are having good determination coefficients respectively 0.96 and 0.99. It is tempting to say that those two expressions should be used to predict the indoor comfort temperature from the monthly mean outdoor temperature particularly because they were very close to the expression given by Humphreys [15]. The expression given by Auliciem and de Dear is found from data covering naturally ventilated buildings and airconditioned buildings. Differences of up to 6 8C were found between the predictions made using the correlation found from our survey’s results and the correlation given by de Dear and Auliciem in Ref. [11]. This difference between both correlations is important especially for low values of the temperature. This difference can be explained by the fact that Auliciem’s correlation takes in consideration both kinds of buildings. From the other side, Nicol found in two different surveys that he has organized in Pakistan [23,24], two correlations very close to the one found by Auliciem despite of the fact that in Pakistan, the buildings are in most of the cases naturally ventilated. The correlation found from de Dear and Brager method seems more robust than the one found according to Griffiths method, but for low outdoor temperatures, the comfort temperature found from Griffiths sounds more plausible. In fact, for an outdoor temperature of 8 8C, the de Dear and
962
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963
Brager correlation predicts a comfort temperature of 12.3 8C which is in our opinion very low unless people wear very heavy clothes which is very restricting for their activity.
8. Conclusion Building standards have been based on fixed comfort temperatures found from tests held in climatic chambers. Those standards assume that the indoor temperature is fixed to a set value and controlled by heating and air conditioning systems. In Tunisia and all North African countries, the heating and air conditioning systems, in case they exist, are not used continuously. Thus, the indoor temperature is fluctuating. The thermal sensation of the building occupants is the only controller of the ventilation, the heating or the cooling of the building. Unlike the conventional thermal regulations which are based on energy consumption, the special feature of the future Tunisian thermal regulation is related to the fact that it must ensure a minimum level of thermal comfort when the building is free running without any heating or cooling system. This study has shown that the Tunisian population has a large potential of adaptation to climate and seasons. Correlation between clothing and temperature relates such potential of adaptation. The comfort temperature can be correlated to the monthly mean outdoor temperature. Such concept can be used to design comfortable buildings. The comparison between the comfort temperature and the maximum and minimum outdoors temperatures can help designer to judge whether passive-heating and cooling techniques are appropriate for the climate under consideration. The relationship between the indoor comfort temperature and the range of outdoors temperatures shows whether for example night ventilation can be or not a way to keep the building cool in summer and help the designer to select the appropriate thermal capacity for the building. Our study suggests that in Tunisia, the thermal comfort temperature can be calculated from one of the expressions: T c-Griffiths ¼ 0:518T o-Avg þ 10:35; T c-Brager ¼ 0:680T o-Avg þ 6:88 where Tc-Griffiths (8C) is the comfort temperature calculated using the Griffiths method, Tc-Brager (8C) is the comfort temperature found according to de Dear and Brager Method, To-Avg (8C) is the monthly mean outdoors temperature. The correlations found in the case of our study are in good agreement with those found in the literature particularly the one found by Humphreys. For buildings equipped with heating and air conditioning systems, a variable indoors temperature has to be taken according to the comfort temperature calculated from the
above equation. This is in agreement with the common way people used to run their heating or cooling systems. Such method of running will allow an important energy saving compared with the existing standards. Our study is a first step towards a thermal comfort standard adapted for the Tunisian conditions. In order to confirm these results, further surveys have to be carried out to validate the results on a larger population.
Acknowledgments The survey has been financed by the ‘‘Agence Nationale pour les Energies Renouvelables’’ in Tunisia and was supported by the British Council. Dr. Fergus Nicol and Rev. M.A. Humphreys have contributed by their directives and advise to the selection of the survey instruments, the preparation of the questionnaires and the data processing. They are gratefully acknowledged for all their support.
References [1] ASHRAE: Standard 55, Thermal environmental conditions for human occupancy, American Society of Heating Refrigeration and Air Conditioning Engineers, Atlanta, GA, USA, 1992. [2] A. Auliciem, R. de Dear, Air conditioning in Australia. I. Human factors, Architectural Science Review 29 (1978) 67–75. [3] A. Auliciems, Towards a psychophisiological model of thermal perception, International Journal of Biometeorology 13 (1981) 147–162. [4] A. Auliciems, Psycho-physiological criteria for global zones of building design, in: Proceedings of the Ninth International Society of Biometeorology Conference, Stuttgart, Hohenheim, 1983. [5] A. Auliciems, R. de Dear, Air conditioning in Australia: human thermal factors, Architectural Science Review 29 (1986) 67–75. [6] C. Bouden, N. Ghrab-Morcos, J.F. Nicol, M. Humphreys, A thermal comfort survey in Tunisia, in: Proceedings of the Second European Conference on Energy Performance and Indoor Climate in Buildings, vol. 2, Lyon, France, 1998, , pp. 491–496. [7] C. Bouden, N. Ghrab-Morcos, Le confort thermique dans les baˆ timents tunisiens: Re´ sultats d’une enqueˆ te, rapport interne publie´ a` l’ENIT, TN, 1999. [8] G.S. Brager, R.J. de Dear, Developing an adaptive model of thermal comfort and adaptation, in: ASHRAE Technical Data Bulletin, vol. 14, No. 1, ASHRAE, Atlanta, USA, 1998. [9] J.F. Bush, A tale of two populations: thermal comfort in air conditioned and naturally ventilated offices in Thailand, Energy and Buildings 18 (1992) 235–249. [10] R. de Dear, B.G. Schiller, D. Cooper, Developing an adaptive model of thermal comfort and preference, ASHRAE Transactions 104 (1) (1988), SF-98-7-3: 4106, RP 884. [11] R. de Dear, A. Auliciem, A field study of occupant comfort and office environments in hot-humid climate, Final report ASHRAE RP-702, 1993. [12] R.J. de Dear, G.S. Brager, Towards an adaptive model of thermal comfort and preference, ASHRAE Transactions 104 (1) (1998). [13] P.O. Fanger, Thermal Comfort, Danish Technical Press, Copenhagen, 1970. [14] I. Griffiths, Thermal comfort studies in buildings with passive solar features: field studies report to commission of the European Community, ENS 35909, UK, 1990.
C. Bouden, N. Ghrab / Energy and Buildings 37 (2005) 952–963 [15] M.A. Humphreys, Outdoor temperatures and comfort indoors, Building Research and Practice 6 (2) (1978) 92–105. [16] M.A. Humphreys, J.F. Nicol, J.F. Nicol, et al. An Adaptive Temperature Guidelines for UK Office Temperatures. Standard for Thermal Comfort: Indoor Air Temperature Standards for the 21st Century, E & F Spon, London, UK, 1995. [17] M.A. Humphreys, J.F. Nicol, Outdoor temperature and indoor thermal comfort: rising the precision on the relationship for the 1998 ASHRAE database of fields studies, ASHRAE Transactions 206 (2) (2000) 485– 492. [18] ISO, International Standard 7730, Moderate Thermal Environments: Determination of PMV and PPD Indices and Specification of the Conditions for Thermal Comfort, International Organisation for Standardisation, Geneva, 1994. [19] D.A. Mc Intyre, Indoor Climate, Applied Science Publisher Ltd., London, 1980. [20] K.J. Mc Cartney, J.F. Nicol, Developing an adaptive control algorithm for Europe, Energy and Buildings 34 (2002) 623–635. [21] K.W.H. Mui, W.T.D. Chan, Adaptive comfort temperature model of air conditioned buildings in Hong Kong, Building and Environment 38 (2003) 837–852. [22] J.F. Nicol, A. Auliciem, A survey of thermal comfort in Pakistan toward new indoor temperature standards, Final Report to the Over-
[23]
[24]
[25] [26]
[27]
[28]
[29]
963
seas Development, Administration Published by Oxford Brookes University, School of Architecture, UK, 1994. J.F. Nicol, I. Raja, A. Alauddin, Thermal comfort in Pakistan II-toward new indoor temperature standards, Report to the Overseas Development Administration, Oxford Brookes University, UK, 1996. J.F. Nicol, I.A. Raja, Thermal comfort and posture. Exploratory studies in the nature of adaptive thermal comfort, Oxford Brookes University, School of Architecture, 1996. ISBN: 1-89940403-1. J.F. Nicol, S. Roaf, Pioneering new indoor temperature standards: the Pakistan project, Energy in Buildings 23 (3) (1996) 169–174. J.F. Nicol, Thermal comfort and temperature standards in Pakistan, in: J.F. Nicol, M. Humphreys, O. Sykes, S. Roaf (Eds.), Standards for Thermal Comfort–Indoor Air Temperature Standards for the 21st Century, E & F Spon, London, 1995, 149–157. J.F. Nicol, I.A. Raja, A. Alauddin, N.G. Jamy, Climatic variations on comfortable temperature: the Pakistan projects, Energy and Buildings 30 (1999) 261–279. K.C. Parsons, Human Thermal Environments: The Effects of Hot, Moderate and Cold Environments on Human Health, Comfort and Performance, Taylor and Francis, 1993. D.M. Rowe, Activity rates and thermal comfort of office occupants in Sydney, Journal of Thermal Biology 26 (2001) 415–418.