Effect of thermal adaptation on seasonal outdoor thermal comfort

Authors


Abstract

Thermal perceptions and preferences of individuals outdoors cannot be entirely explained by the energy balance of the human body. They are also affected by psychological and behavioural factors or the so-called thermal adaptation. To examine the effect of thermal adaptation on seasonal outdoor thermal comfort, 1644 interviews with concurrent micrometeorological measurements were conducted outdoors in central Taiwan. Results indicate a deviation of 1.3 °C standard effective temperature (SET*) in neutral temperatures between hot and cool seasons, and a deviation of 1.8 °C SET* in preferred temperature between hot and cool seasons. Additionally, although subjects' temperature and sunshine preferences were highly correlated with SET*, they diverged between seasons for identical SET* exposures in the two seasons. Analysis reveals that people's thermal perceptions were strongly related to the air temperature (Ta) and mean radiant temperature (Tmrt), but not significant to air speed and air humidity. These results demonstrate that thermal adaptation markedly influences seasonal outdoor thermal comfort, knowledge of which may be useful in the planning and design of outdoor environments in hot–humid regions. Copyright © 2010 Royal Meteorological Society

1. Introduction

Many leisure and recreational activities are conducted in outdoor environments, exposing participants directly to the weather. Several studies conducted in a wide range of climate zones have indicated that the number of individuals who use outdoor spaces is related to air temperature (Ta), mean radiant temperature (Tmrt) and other thermal comfort indices (Nikolopoulou et al., 2001; Nakano and Tanabe, 2004; Thorsson et al., 2004; Eliasson et al., 2007; Thorsson et al., 2007a; Lin, 2009). As many outdoor microclimates are partially determined by the built environment, including, for example, ground surface cover (Lin et al., 2007; Wong et al., 2007), canyon geometry (Ali-Toudert and Mayer, 2006; Kolokotroni and Giridharan, 2008; Giridharan and Kolokotroni, 2009; Lin et al., 2010), and many recreational activities of considerable commercial value which are performed outdoors, thermal comfort in outdoor environments is therefore an important research area (Spagnolo and de Dear, 2003; Lin and Matzarakis, 2008).

Some field studies have focused on the thermal comfort of samples of individuals in outdoor or semi-outdoor environments (Ahmed, 2003; Spagnolo and de Dear, 2003; Nakano and Tanabe, 2004; Oliveira and Andrade, 2007). In Taiwan, a hot–humid region, Hwang and Lin (2007) conducted a thermal comfort survey as part of the field investigation of outdoor thermal comfort project (FIOT). They analysed the neutral temperature, preferred temperature and range of acceptable temperatures for indoor, outdoor and semi-outdoor samples, and found significant variation in thermal comfort responses between these different settings.

The comfort literature indicates that thermal perceptions and preferences cannot be fully explained in terms of the energy balance of the human body: they are also affected by various psychological and behavioural factors, including thermal experience, comfort expectations, perceived thermal control, culture and duration of exposure (Paciuk, 1990; Brager and de Dear, 1998; Nikolopoulou et al., 2001; Nikolopoulou and Steemers, 2003). These contextual factors are collectively referred to as ‘thermal adaptation’. The variance of people's thermal comfort with season is a significant example of thermal adaptation, which is related to subjective experience of, and expectations for, thermal comfort. To examine the effect of thermal adaptation on seasonal outdoor thermal comfort, the following three issues must be clarified. The first is the variation of basic thermal comfort requirements with the seasons. For example, Spagnolo and de Dear's investigation in humid subtropical Sydney, Australia, revealed that subjects' neutral temperature shifted from 23 °C in winter to 33 °C in summer as the background outdoor Ta shifted from 12 °C in winter (minimum) to 43 °C in summer (maximum) (Spagnolo and de Dear, 2003). Outdoor field experiment by Nikolopoulou et al. in temperate Cambridge, UK, observed subjects' neutral temperature shifting from 10 °C in winter to 30 °C in summer as the background outdoor Ta shifted from 1 to 29 °C (Nikolopoulou et al., 2001). The present study investigates whether these large changes in neutral temperate with the changing seasons also occur in hot and humid regions, such as Taiwan, where seasonal drift in Ta is relatively small (approximately 10 °C throughout the year).

The second issue requiring clarification concerns whether thermal indices accurately reflect people's actual thermal preferences. Thermal indices, such as standard effective temperature (SET*) (Gagge et al., 1986) and its outdoor variant, OUT_SET* (de Dear and Pickup, 1999; Pickup and de Dear, 1999; Spagnolo and de Dear, 2003) and physiologically equivalent temperature (PET) (Mayer and Höppe, 1987; Höppe, 1999) are based on the climate-chamber analyses of the human energy balance and integrate the effects of Ta, vapour pressure (VP) or relative humidity (RH), Tmrt and air speed (v). However, as noted earlier, contextual factors induce thermal expectations that are specific to each setting, so energy-balance indices may not be universally applicable across all contexts. Accordingly, people may well have divergent thermal perceptions/preferences when they are exposed to different contexts, despite having identical thermal balances as indicated by the heat-balance comfort indices.

The third thermal comfort issue that requires clarification is the relative influence of the various heat-balance parameters on thermal perception. Humphreys and Nicol (2004) employed data collected in the smart controls and thermal comfort (SCATs) project (Nicol and McCartney, 2001; McCartney and Fergus Nicol, 2002) and in American Society of Heating, Refrigerating and Air-conditioning Engineers (ASHRAE) RP-884 (de Dear, 1998; de Dear and Brager, 1998), and calculated the relative contribution of each physical microclimatic parameter to thermal perception by performing multivariate analyses on the ASHRAE 7-point thermal sensation scale. They found that indoor Ta dominated (78–83%) thermal sensation, followed by outdoor temperature (9–10%), with RH, v, clothing insulation and metabolic rate explaining less than 5% of the variance in thermal sensation. Hwang et al. (2009) conducted a thermal comfort survey in a climate chamber in Taiwan, using the orthogonal array method of data analysis. They also found that Ta dominated (69%) thermal perception, whereas the other aforementioned parameters only explained less than 11% of the variance in thermal sensation. The present study will also examine the contribution of physical microclimatic parameters to people's subjective thermal perception in the hot and humid outdoor setting of Taiwan.

In summary, the goals of this study are as follows:

  • 1.quantify subjects' basic thermal comfort requirements in different seasons in the hot–humid climate of Taiwan;
  • 2.determine whether subjects' thermal preferences can be effectively explained by the heat-balance indices alone in the different seasons of Taiwan;
  • 3.determine the relative contributions of heat-balance parameters to thermal perception in an outdoor environment in different seasons.

The results of these analyses will be used to assess the effects of thermal adaptation on seasonal outdoor thermal comfort in the hot–humid outdoor settings of Taiwan.

2. Analysis of the FIOT project data

The data analysed in this study were taken from the FIOT project (Hwang and Lin, 2007) in which both physical microclimatic measurements were made concurrently with questionnaire surveys in various indoor, semi-outdoor and outdoor spaces in central Taiwan, including Taichung, Yunlin and Chiayi (Figure 1). A total of 8077 sets of data were collected from the winter of 2004 to the summer of 2005. Figure 2 presents photographs of the locations of the field experiments in Taichung and Yunlin. A total of 1644 interviews from the FIOT database, drawn from the outdoor surveys are analysed in this paper.

Figure 1.

Maps of Taiwan and the locations of field experiments

Figure 2.

Photos of locations of field experiment in Taichung (left) and Yunlin (right). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

The physical microclimatic parameters considered in the analysis are Ta, RH, VP, v, global solar radiation (G) and globe temperature (Tg), all measured at the height above ground corresponding with the subjects' waist (1.1 m) using instruments compliant with the ISO 7726 standard (ISO, 1998). Tmrt was estimated from Ta, v and Tg, using the equations in ISO (ISO, 1998; Thorsson et al., 2007b) and the RayMan model (Matzarakis et al., 2007), which has been validated elsewhere (Lin et al., 2010). VP was calculated by multiplying saturation VP at the Ta by RH. A questionnaire survey was administered during the measurements of physical microclimate. The first part of the questionnaire collected demographic information such as age and gender, and data concerning metabolic rate (activity levels) and clothing insulation worn by subjects (garment checklist). The second section asked subjects to rate their current thermal comfort level using standardised scales. Thermal sensation was rated on the ASHRAE 7-point thermal sensation vote (TSV): − 3, cold; − 2, cool; − 1, slightly cool; 0, neutral; 1, slightly warm; 2, warm; 3, hot). Thermal preferences were assessed using the McIntyre preference scale (right now I prefer ‘cooler,’ ‘no change’ or ‘warmer’), whereas thermal acceptability was directly assessed with a binary item ('acceptable' or ‘unacceptable’). Subjects were also asked to rate their instantaneous sensation of, and preference for, wind and sunshine. Wind/sunshine sensations were rated on a 7-point scale from − 3 for ‘very weak’ to 3 for ‘very strong’ (0 = ‘neutral’), and preferences were rated on a 3-point scale with, ‘I would prefer wind/sunshine to be weaker’, ‘I prefer no change in wind/sunshine’ and ‘I would prefer wind/sunshine to be stronger’. Sensation of, and preference for humidity were similarly rated.

The physical microclimatic parameters of Ta, RH, Tmrt and v were integrated into the physiologically relevant index SET* (Gagge et al., 1986). According to the meteorological data from Taichung City in 1970–2000, average Ta peaks in July at 28.5 °C (maximum temperature, 33 °C), and mean Ta is lowest in January at 16.2 °C (minimum temperature, 12.4 °C). Average RH during a year is 70–80%. Meteorological data reveal that summers are hot and winters are mild in central Taiwan. As Ta is low only from December to February, this period is designated herein as the ‘cool season’, whereas the remaining months (March to November) are referred to as the ‘hot season’.

3. Basic thermal comfort requirements observed in different seasons

3.1. Clothing

Thermal adaptation theory suggests that people take spontaneous behavioural actions to make themselves feel thermally comfortable. The most important and most convenient behaviour is to adjust clothing insulation. Analysis of observed clothing levels in the hot and cool seasons for each SET* interval was performed and mean values of clothing insulation level (clo) were estimated from individual garment checklists published in ISO 7730 (ISO, 2005). Figure 3 plots the relationship between subjects' clothing and concurrent Ta in both the hot and cool seasons of central Taiwan. The graph indicates a trend for the amount of clothing insulation worn by the subjects to decline as Ta increased. In the cool season, the mean clothing insulation value increased to 1.1 clo at a mean Ta of 14 °C. In the hot season, the amount of clothing decreased to 0.5 clo at the corresponding Ta of 28 °C. The amount of clothing worn remains around 0.5 clo when Ta > 28 °C, perhaps because 0.5 clo represents the wearing of only a short-sleeve T-shirt and light pants, and any further removal of clothing is limited by social norms. Over a range of temperatures, Ta = 22–28 °C, people wear a little more in the cool season (0.71 mean clo) than in the hot season (0.60 mean clo) when exposed to the same Ta of 22–28 °C, suggesting a seasonal offset in clothing insulation regardless of instantaneous temperatures.

Figure 3.

Clothing insulation (clo units) worn by subjects for each Ta bin. The box represents the mean value and the whiskers represent the 95% confidence interval. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

3.2. Observed neutral temperatures

The temperature coinciding with the central thermal sensation (TSV = 0) in the ASHRAE 7-point scale is referred to as the neutral temperature; the temperature at which people feel neither cool nor warm (Fanger, 1972). The general approach to obtain neutral temperature is to calculate the mean TSV (MTSV) within each temperature interval, and then fit a linear regression function between MTSV and SET* (de Dear and Fountain, 1994). The width of temperature bins used in this analysis was 1 °C SET*. Figure 4 plots the MTSV observed in each SET* bin, for both hot and cool seasons. The optimally fitting linear equations are:

equation image(1)
equation image(2)
Figure 4.

Correlation between MTSV and SET* in central Taiwan's hot and cool seasons. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Substituting MTSV = 0 into Equations (1) and (2) yields the neutral temperatures for subjects; 29.3 and 28 °C SET* in hot and cool seasons, respectively. A difference of 1.3 °C SET* exists between the neutral temperatures in the hot and cool seasons, revealing a moderate effect of seasonal adaptation on thermal comfort. This effect is not the result of behavioural adaptations such as clothing adjustments as clo is one of the parameters included in the calculation of SET*, so the explanation of seasonally offset neutralities is more likely to be a psychological adaptation such as a differentiation of seasonal comfort expectations.

The difference between the neutral temperatures in hot and cool seasons in Taiwan, i.e. 1.3 °C SET*, is much smaller than that of UK and Australia, i.e. 10 and 20 °C, respectively, as mentioned in Section 1. The comparative result may be due to that the annual variation of Ta of Taiwan (about 15 °C) is much smaller than that of UK and Australia, i.e. 28 and 31 °C, respectively. These results reveal that the difference in neutral temperature in hot and cool seasons in a certain country is related to its annual Ta variation.

3.3. Observed preferred temperature

Preferred temperature can be defined as the thermal environmental condition under which individuals prefer neither warmer nor cooler temperatures (Fanger, 1973). Although the neutral temperature is the temperature which the subjects collectively describe as neutral, preferred temperature is the temperature people actively prefer, and this subtle semantic difference opens up the possibility for neutral and preferred temperatures to be different. Therefore, this investigation derived preferred temperatures in both seasons from thermal preferences (McIntyre's scale) given in response to questionnaires, verifying the effect of seasonal expectations on thermal comfort.

Probit analysis (Ballantyne et al., 1977) was utilised to calculate preferred temperature based on stated preferences for warmer or cooler temperatures. Preferences were assessed within bins that correspond to 1 °C SET* intervals; the percentage of preferences is calculated within each bin and fitted to SET* using a logistic (S-shaped) curve model with the probit function. The SET* at which both ‘prefer warmer’ and ‘prefer cooler’ curves intersect represents the temperature at which a majority of the sample of subjects preferred neither a cooler nor a warmer temperature. This is the operational definition of preferred temperature (de Dear and Fountain, 1994).

According to Figure 5, the preferred temperatures in central Taiwan's hot and cool seasons were 28.5 and 26.7 °C SET*, respectively, with a difference of 1.8 SET*, suggesting the effects of thermal adaptation in the different seasons. Accordingly, the differences between neutral and preferred temperature are 0.8 °C in the hot season and 1.3 °C in the cool season. Although these differences are small, the comparative results still somewhat provide some evidence of semantic artefacts.

Figure 5.

Preferred temperatures in hot and cool seasons. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

4. Temperature, humidity, sunshine and wind preferences

As stated in Section 1, people may have different thermal perceptions/preferences in different contexts despite the thermal comfort indices indicating thermal parity. This study seeks to clarify whether people's thermal preferences can be completely interpreted in terms of thermal indices, or whether their preferences are determined by the combination of physical microclimatic parameters and psychological factors.

To explore this issue within the FIOT database, all datasets were sorted into bins, each corresponding to a 1 °C SET* interval; Figures 6–9 plot the subjects' thermal preferences within each bin, together with the corresponding mean values of the relevant physical microclimatic parameters for the bins. For example, Figure 6(a) plots the percentages of preferred temperatures (prefer warmer, cooler) and the mean Ta for each SET* bin in the hot season, whereas Figure 6(b) does the same for the cool season. Similarly, Figures 7–9 plot the humidity, sunshine and wind preferences, respectively.

Figure 6.

Subjects' preference for temperature versus mean Ta for each SET* bin in the hot (a) and the cool seasons (b). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Figure 7.

Subjects' preference for humidity versus mean VP for each SET* bin in the hot (a) and the cool seasons (b). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Figure 8.

Subjects' preference for sunshine versus mean Tmrt for each SET* bin in the hot (a) and the cool seasons (b). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Figure 9.

Subjects' preference for wind versus mean air speed for each SET* bin in the hot (a) and the cool seasons (b). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Figure 6(a) and (b) presents the subjects' preference for temperature versus mean Ta for each SET* bin in both the hot and cool seasons. In the hot season, as SET* increases, the percentage who ‘prefer warmer’ decreases and the percentage who ‘prefer cooler’ increases, reaching 100% at about 44 °C SET*. In the cool season, although the percentage preferring to feel warmer declines as SET* increases, the percentage who ‘prefer cooler’ does not increase significantly, remaining generally between 30 and 50% once SET* exceeds 25 °C. A comparison of the two seasons indicates that the percentage who ‘prefer cooler’ in the hot season markedly exceeds that in the cool season for high SET* bins (such as the 34–39 °C SET* bins). The percentage who ‘prefer cooler’ in the hot season is greater than that in cool season because of the distribution of mean Ta in both seasons. When people are exposed to 34–39 °C SET*, the mean Ta is 30–32 °C in the hot season, which is 4 °C higher than in the cool season (25–28 °C). The high SET* bins (34–39 °C SET*) occurring in winter with low Ta may be due to an elevated radiant load (Tmrt) for observations falling in those bins. As people may be more tolerant of high Tmrt in the cool season compared with the hot season, the percentage that ‘prefer cooler’ in the hot season is higher than in the cool season. This comparison reveals that the subjects' temperature preferences can be interpreted in terms of SET* for each season. However, their temperature preferences vary with the seasons, despite the thermal comfort indices' being identical.

Figure 7(a) and (b) displays the subjects' preference for air humidity versus mean VP for each SET* bin in both the hot and cool seasons. In the hot season, subjects' preference for humidity is slightly affected by SET* or the mean VP, whereas in high SET* bins (exceed 35 °C SET*). In the cool season, subjects preferred drier conditions only when the SET* was low (<24 °C). Despite this trend, it is worth noting that the actual VP was almost constant (15–20 hPa) across all SET* bins. A possible explanation is that subjects preferred drier because they felt that it would help to eliminate their overall discomfort, based on their previous experience. Analytical results indicate that the humidity preference cannot be explained in terms of SET*, and that the preferred humidity is related only to the mean VP when VP is high in the hot season.

Figure 8(a) and (b) plots the subjects' preference for sunshine intensity versus actual measurements of the mean Tmrt for each SET* bin in both the hot and cool seasons. In both hot and cool seasons, as SET* and Tmrt increases, the percentage who ‘prefer stronger sunshine’ declines, and therefore the percentage who ‘prefer weaker sunshine’ increases. It should be noted that the cool season sample of subjects preferred more sunshine than their hot season counterparts experiencing the same levels of SET*.

Figure 9(a) and (b) shows the subjects' preference for wind versus mean air speed (v) for each SET* bin in both the hot and cool season surveys. In the hot season, as overall heat loads (SET*) on the subjects increased, the percentage subjects preferring stronger wind also increased, reaching 80% at > 43 °C SET*. Despite this trend, it is worth noting that the actual wind speed was almost constant (0.8–1.2 m/s) across all SET* bins. A possible explanation is that subjects preferred more wind because they felt that it would help to eliminate their overall discomfort, based on their previous experience. In the cool season, the wind preference results were similar to those in the hot season. However, the percentage of subjects who ‘prefer stronger wind’ at high SET* (34–38 °C SET*) in the cool season was only 35%, which is significantly lower than in the hot season. This finding suggests that wind preferences are contextualised to season, and this context may override the body's instantaneous heat-balance status as the driver for wind preference.

In summary, the results in Figures 6–9 reveal that people have different thermal preferences in different contexts, despite their having identical values of SET*. The results also suggest that the thermal preferences for outdoor people cannot be completely explained by thermal indices based on the energy balance of the human body. Moreover, the conditions of temperature, humidity, sunshine and wind that are preferred by people outdoors are affected not only by the prevailing values of related physical microclimatic parameters but also by their experience and expectation in different seasons, which in fact provides the evidence of thermal adaptation in the outdoor environment.

Another explanation why peoples' preferred wind speed and air humidity are not closely correlated to actual wind speed and air humidity may involve the climate in Taiwan. When Kolokotroni and Giridharan discussed the urban heat island in London, they divided their observations into three wind spectra (10, 5 and 2.5 m/s). In this investigation, the measured wind speed was normally 0.8–1.2 m/s, which is lower than in London. Additionally, the average RH during a year is 70–80% in central Taiwan. As wind speed and RH are stable throughout the year, people's thermal preferences are more strongly related to a combination of Ta and Tmrt because these variables vary significantly throughout the year.

5. Relative contributions of physical microclimatic parameters to thermal perceptions

Analysis of variance (ANOVA) was employed here to evaluate the relative contributions of physical microclimatic parameters to people's thermal perceptions in an outdoor environment. In this analysis, the dependent variable is TSV and the independent variables are the three physical microclimatic parameters—operative temperature (To, arithmetic average of the Ta and Tmrt), VP and v. Firstly, the subjects are split into groups by applying a certain boundary values of each parameter, and the numbers of subjects in each group should be equal for ANOVA analysis. If the subjects are divided into too many groups, then the thermal sensations of the few subjects in each group may be not representative. If too few groups are used, then significant differences among groups may not be evident. After several trials, the most significant variances between groups were found when the subjects were divided into four groups for each parameter. Hence, quarterly boundaries were applied in the ANOVA analysis herein. In the hot seasons, the quartile boundaries are 29.1, 33.2 and 39.9 °C for To, 16.9, 26.9 and 32.3 hPa for VP, and 0.7, 1.0 and 1.3 m/s for air speed, whereas in the cool seasons the quartile boundaries are 22.8, 27.2 and 33.7 °C for To, 16.3, 17.8 and 19.0 hPa for VP, and 0.8, 1.0 and 1.3 m/s for air speed. After sorting into quartiles, univariate analysis was performed using statistical software.

Table I presents the relative contributions of the physical microclimatic parameters to the ‘explained’ variance (TSV in hot and cool seasons. The contribution of To to variance in TSV was highly significant in both hot and cool seasons, whereas those of VP are moderately significant and v was negligible. The comparative results demonstrate that people's thermal perception is strongly affected by the To (which is the combination of Ta and Tmrt), consistent with the results in Figures 6 and 8. Notably, VP is twice as important at explaining variance in TSV during the hot season (23.9%) than in the cool season (12.1%). This finding is consistent with the accepted wisdom that humidity is only relevant to thermal comfort when skin wettedness is high—resulting from either elevated metabolic or environmental heat loads on the body. The comparative results reveal the subjects' thermal comfort variation in different seasons, reflecting the climate characteristic in hot-humid Taiwan.

Table I. Relative contribution of micrometeorological parameters to the explained variance in TSV
 Relative contribution (%)
 Hot seasonCool season
  • *

    p-value < 0.05.

  • **

    p-value < 0.001.

Operative temperature75.7**87.3**
Vapour pressure23.9*12.1*
Wind speed0.40.5

It should be noted that air humidity and wind speed are stable throughout the year in Taiwan. This may also be the reason why the relative contributions of VP and v to the TSV are low.

6. Conclusions

This work utilises the 1644 interviews conducted in the FIOT projects to analyse outdoor thermal comfort. The basic thermal comfort requirement varies with season. The neutral temperatures are 29.3 and 28 °C SET* in the hot and cool seasons, respectively, and the preferred temperatures are 28.9 and 25.4 °C SET*, revealing the effect of seasonal psychological adaptation on thermal comfort. As the seasonal difference in neutral and preferred temperatures is different, with preferred temperature showing the bigger seasonal offset, this raises a question about factors other than seasonal adaptation playing a role. Other authors have referred to the ‘semantic artifact’ hypothesis when such discrepancies have been noted between preferred and neutral temperatures (de Dear and Brager, 1998). The hypothesis suggests that people may describe their preferred thermal state with different adjectives in different seasons. In hot summer conditions, it is conceivable that people would describe their preferred thermal state as ‘slightly cool’ while in winter the preferred thermal state may be more accurately described as ‘slightly warm.’ According to this hypothesis, one would expect preferred temperature in the hot season to be cooler than the corresponding neutral temperature, and this was indeed the case in the present study in Taiwan, although only by a small amount (0.8 °C). Similarly, the semantic artefact hypothesis suggests that cool season preferences would be warmer than the corresponding neutral temperature, but this was not borne out in the present study—preferred temperature in Taiwan's cool season was much cooler than the neutral temperature registered in the same season (1.3 °C cooler). Therefore, it appears as if the semantic artifact hypothesis may not be applicable to these findings in central Taiwan, probably because the contrast among seasons in Taiwan is not as great as it is in the mid-latitude climate zones for which the hypothesis was originally proposed. As the winter in central Taiwan feels mild, especially when people are exposed to strong solar radiation, they still prefer lower temperatures outdoors in winter. The relatively modest difference neutral temperatures (1.3 °C SET*) between hot and cool seasons in Taiwan is less than that in UK or Australia, probably because Ta is more stable over the year in Taiwan.

The results of an analysis of subjects' thermal preferences indicated that the temperature and sunshine preferences can broadly be explained using the heat-balance index SET*. However, people have different preferences for Ta when experiencing identical heat-balance states (SET*) in the two different seasons. The above results indicate that thermal indices such as SET* can serve only as a reference in the evaluation of subjects' preferences concerning outdoor physical microclimatic parameters. A combination of physical microclimatic parameters, past thermal experiences (thermal history) and expectations in different seasons may affect their thermal preferences in a more complex way than the heat-balance indices such as SET* can handle.

Concerning contribution of physical microclimatic parameters to thermal perceptions, comparative results reveal that people's thermal perceptions are strongly affected by To, which is the mean value of Ta and Tmrt, moderately affected by VP but not significantly by v. Furthermore, the result is related to Taiwan's climate characteristic and subjects' thermal comfort variation in different seasons.

This investigation not only confirms the effect of thermal adaptation on seasonal outdoor thermal comfort, but also demonstrates the limitation of thermal indices based exclusively on a heat-balance analysis of the human subjects in predicting their thermal preferences. The physical microclimatic parameters that dominate subjects' thermal perceptions are also explained in terms of the local weather and people's experiences.

By elucidating outdoor occupants' thermal comfort, the results of this study may contribute to the planning and design of outdoor environments in hot–humid regions, supporting the use of outdoor spaces and increasing their occupants' satisfaction. For example, the results suggest that a reasonable hot season design temperature range for occupants of outdoor recreational and sporting facilities in central Taiwan is about 29 °C operative temperature. Although the present study did not attempt to fit an acceptability range around preferred temperatures, on the basis of earlier comfort research (de Dear and Brager, 1998), ± 3 °C is not an unreasonable estimate, suggesting a design range of 26–31 °C in Taiwan's hot season. The corresponding estimate of Taiwan's cool season acceptable temperature range is 26.7–3.0 to 26.7 + 3.0, or simply 23.7–29.7 °C. By overlaying these thermal comfort ranges on central Taiwan's typical reference years of meteorological data, architects and engineers have a rational basis for determining whether their proposed outdoor and semi-outdoor facilities would be acceptable to occupants, and if they will not be, ameliorative strategies may be applied to the designed environment. For example, if meteorological observations will fall above the acceptable range in the hot season, more shade should be provided to the design so that Tmrts are brought down to reasonable levels. Alternatively, elevated air speeds may be achievable by adjustments to the design, and these may be effective in brining unacceptably high temperatures back into the acceptable range of 26–31 °C in summer.

Acknowledgements

The authors thank the National Science Council (NSC) of Taiwan for financially supporting this research under contract no. 98-2221-E-150-063.

Ancillary