Effects of various environmental grades on thermal comfort in semi-outdoor spaces

Abstract

The introduction of semi-outdoor spaces to urban spaces and offices has increased in recent years. In previous studies, environmental grade has been defined as the level of control for the thermal environment set for an architectural space based on the use, where “semi-outdoor environment” is an environmental grade division that exists as a gradation between indoors and outdoors. This study aims to clarify effects of various environmental grades on thermal comfort. Field surveys were conducted in five semi-outdoor spaces in Tokyo from 9 to 25 September 2021, consisting of thermal environment measurement, questionnaires about thermal comfort and impression of surveyed spaces and real-time sensation voting. By impression evaluation and comparison of the regression curves of standard new effective temperature (SET*) against percentage of “uncomfortably warm,” the surveyed spaces were classified into two categories: “outdoor tendency” and “indoor tendency.” The difference in thermal comfort for the diverse semi-outdoor environment could be evaluated by the difference in spatial characteristics such as “sense of weight” of the top surface in addition to the difference in environment control. By regression analysis of real-time sensation voting result using hierarchical Bayesian approach, environmental grade was clarified to affect the correlation between pleasantness/unpleasantness and air velocity.

Nomenclature
• a_k
• Absorption coefficient of irradiated body surface area for short wavelength radiation (= 0.7) [−]
• Diffuse
• Heat received by the human body from diffuse solar radiation [W/m²]
• Direct
• Heat received by the human body from direct solar radiation [W/m²]
• I*
• Normal surface direct solar radiation separated from S↓ [W/m²]
• IdH
• Diffuse solar radiation on the horizontal surface separated from S↓. [W/m²]
• L↓
• Downward long wavelength radiation [W/m²]
• L↑
• Upward long wavelength radiation [W/m²]
• Longwave
• Heat received by the human body from longwave radiation [W/m²]
• MRT_2dir
• Mean radiance temperature calculated from long and short wavelength radiation in the two directions upward and downward [°C]
• S↓
• Downward short wavelength radiation [W/m²]
• S↑
• Upward short wavelength radiation [W/m²]
• ε_p
• Emission coefficient of the human body (=0.97) [−]
• f_p
• Projected area factors of the human body (=0.174) [−]
• σ
• the Stefan-Boltzmann constant (= 5.67 × 10⁻⁸) [W/(m² · K⁴)

Introduction Importance of semi-outdoor spaces

The introduction of semi-outdoor spaces to urban spaces and offices is increasing as a space for interaction among residents and performing simple office tasks.^1,2 Semi-outdoor space is not just limited to being a buffer space but has also diversified to include spaces which can accommodate long duration of stays for the occupants. In the future, the use of such spaces is likely to increase. Furthermore, the introduction of semi-outdoor space that creates an unsteady and heterogeneous environment by incorporating elements of the outdoor environment may contribute to a reduction of the cooling/heating load, as well as stress relief and a change of pace for those staying in the area.³ In addition, Mihara et al.⁴ showed in a subject experiment that short-term work activities in a semi-outdoor space can be performed without decreasing work performance. Against this background, the importance of semi-outdoor spaces should continue to increase.

Previous on thermal comfort in the semi-outdoor environment

Nakano et al.⁵ defined the level of control for the thermal environment set for an architectural space based on the use as environmental grade, where “semi-outdoor environment” is an environmental grade division that exists as a gradation between indoors and outdoors. The definition of “semi-outdoor space” in this study is an architectural space for which the control level of the thermal environment is classified as semi-outdoor environmental grade. For the semi-outdoor environment where the own adaptation of the occupant to the environment is considered while assuming that the environmental factor is controlled by architectural elements and facilities, evaluation of thermal comfort with a predicted mean vote (PMV) standardized by ISO 7730⁶ is difficult.⁷ Nakano et al.^7–9 examined the impact of behavioral adaptation, such as the reason for stay and clothing adjustment and thermal comfort for the semi-outdoor environment where the thermal environment is controlled differently through the presence/absence of air-conditioning. The result showed that the thermal comfort zone for the semi-outdoor environment was wider than that of the indoor thermal comfort zone (percentage of dissatisfied <20%) indicated by American Society of Heating, Refrigerating, and Air-Conditioning Engineers Standard 55 (ASHRAE Standard 55).¹⁰ Specifically, occupants allow for environmental fluctuations by external disturbances, such as solar radiation and wind, where environmental adaptation by the occupants impact the thermal comfort. Studies by Hwang et al.¹¹ and Spagnolo et al.¹² have also shown that the thermal comfort zone is expanded and neutral temperatures are higher in semi-outdoor environments than in indoor environments. Gamero-Salinas et al.^13,14 have investigated the influence of building form variables on the physical environment. However, few studies conducted subject experiments on thermal comfort in semi-outdoor spaces considering spatial characteristics.

Brager et al.¹⁵ classified environmental adaptation into three types: behavioral adaptation in which thermal equilibrium for the body is maintained through conscious or unconscious behavior, physiological adaptation in which the body adapts through genetics and to seasons, and psychological adaptation in which the recognition of thermal environment is adjusted through habituation based on previous experiences and expectations of the environment. Nakano et al.⁷ and de Dear et al.¹⁶ focused on the difference in environmental control methods within psychological adaptation and showed the possibility of expectations for the environment being eased in natural ventilation and non-air-conditioned space, extending the tolerance of the occupant for the thermal environment. Psychological adaptation has an especially high impact on thermal comfort¹⁵; however, not many findings exist that clarify its impact on the semi-outdoor environment.

Previous studies on real-time sensation voting regarding pleasantness/unpleasantness

It is difficult to record the impact of the thermal environment on comfort instantly using voting with a questionnaire every several minutes. Real-time sensation voting devices, ostracon¹⁷ and YUHO,¹⁸ supplement weaknesses in the questionnaire method and make immediate recording of psychological quantities. In this method, compliant and feelings of pleasantness/unpleasantness regarding the thermal environment can be analyzed. Thermal comfort surveys that use real-time sensation voting are useful for the semi-outdoor environment where environmental fluctuation due to external disturbance is notable.

Aim of this study

As such, environmental adaptation of occupants changes according to the shape and usage of the space in the semi-outdoor environment, possibly resulting in an impact on thermal comfort. To design an appropriate thermal environment for multipurpose semi-outdoor spaces, the environmental grade defined by Nakano et al,⁵ semi-outdoor environment, needs to be further subdivided, and the thermal comfort of occupants needs to be understood based on this classification. Additionally, the correlation between the physical environment and pleasantness/unpleasantness for semi-outdoor spaces must be clarified using real-time sensation voting. Here, we aimed to clarify the impact of the difference in environmental grades of semi-outdoor spaces on thermal comfort.

Material and Methodology

In this study, we conducted physical environment measurements, questionnaire surveys, and real-time sensation voting surveys at five semi-outdoor spaces on the Nishiwaseda Campus, Waseda University between 10:00 to 17:00 over 7 days between September 9 and September 25, 2021 [Note 1]. First, we classified surveyed spaces into two environmental grades using a qualitative method that is based on an impression evaluation of spaces. This was followed by a comparison of the relationship between the Standard New Effective Temperature (SET*) and the percentage of uncomfortable vote for each surveyed space to clarify the impact of different environmental grades of semi-outdoor spaces on the thermal environment tolerance of occupants. To discuss the impact of spatial characteristics on the conditions for pleasantness/unpleasantness, we focused on the air velocity with higher temporal variations than other environmental elements. By using the Hierarchical Bayesian approach, we obtained a regression equation for pleasantness/unpleasantness regarding the air velocity and compared the impact of airflow on thermal comfort for each surveyed space.

Surveyed spaces

Figure 1 shows the actual condition of the measurements, questionnaire, and real-time sensation voting surveys at each surveyed space. Five surveyed spaces were selected with the following conditions: architectural space where outdoor environmental elements were incorporated and it is assumed that the space would be occupied by users. Surveyed spaces A, B, C, D, and E were a terrace in the shade of trees with a north–south breeze, a two-story atrium with large glass openings on the east–west surfaces, a courtyard with benches in the shade of trees, a two-story atrium student lounge with a large opening to the east, and a piloti facing the outdoor pathway in a north–south direction, located on the east side of a building, respectively. It is surrounded by other buildings.

Measurement schedule

Figure 2 shows the measurement schedule. Subjects were 30 men and women in their 20s. Clothing condition was half-sleeve shirt, long pants, and shoes, which were prepared by the subjects. To unify the evaluation standard for environmental grades, subjects rested in a chair in pre-rooms that was air-conditioned to about 25°C then spent 10 min in a fixed outdoor point with an exposure to direct solar radiation (“outdoor environment”) and 10 min in the pre-rooms (“indoor environment”). Subsequently, subjects rested in a chair for 30 min in a semi-outdoor space without direct solar radiation or air-conditioning for measurement and questionnaire. Number of subjects and clothing insulation for each measurement day is shown in Table 1. In principle, different subjects participated each day [Note 2].

Measurement items

Table 1 shows physical environment and psychological quantity measurement items. The physical environment near subjects was measured with the movements of subjects by using a mobile measurement cart with measurement devices [Note 3]. The outdoor physical environment was continuously measured at the fixed outdoor points (see Figure 1). To calculate mean radiant temperature (MRT), according to VDI3787,¹⁹ we used a calculation method based on the actual measurements of longwave and shortwave radiation [Note 4]. In this measurement, subjects stayed in semi-outdoor spaces without exposure to direct solar radiation; thus, all measured short wavelength radiation was calculated by assuming diffuse solar radiation. As a psychological quantity [Note 5], we had subjects vote on thermal sensation, thermal comfort, thermal preference, and thermal acceptance every 2.5 min after 5 min of stay (total of 11 times). Dry and wet sense, feeling of sunlight, air movement sensation, and comfort with air movement were voted on once after a 30-min stay by drawing a diagonal line at the relevant position on a subjective scale (see Figure 3) using a questionnaire. A real-time sensation voting survey, discussed below, was conducted at each surveyed space where pleasantness/unpleasantness was voted on using smartphones.

Table 1 Measurement items of physical environment and psychological quantity

[Table omitted. See PDF]

Table 2 shows questionnaire items for the impression evaluation. As for spatial composition, space around subjects was divided into six faces, and subjects evaluated each of four vertical faces (elevation) and the top surface, as well as the total space, using the subjective scale shown in Figure 3. Questionnaire items were prepared around the extent of natural environmental elements that could be sensed, such as sunlight, solar radiation, and air flow. For the evaluation of the thermal environment, we surveyed the similarity of the thermal environment of the surveyed spaces to the outdoors, compared to that of the indoor environment, as found by the subjects. The impression evaluation was performed at fixed outdoor points and the pre-rooms. As for clothing, subjects selected clothing they wore during the measurement from a checklist. The “clo” value of each clothing from ISO 9920²² was added to estimate the clothing value for the combination of clothes each subject wore.

Table 2 Questionnaire contents on impression evaluation

		Questionnaire
Elevation	Sunlight transmittance	The object or shielding is likely to transmit the sunlight to your vicinity
	Solar heat transmittance	The object or shielding is likely to transmit the solar heat to your vicinity
	Breeziness	The object or barrier is likely to let the wind pass through against your vicinity
	Visibility shielding	I feel that my view is obstructed by that object or barrier
	Degree of outdoor	How much do you feel separated from the outdoor environment by the object or barrier?
Top surface	Sunlight transmittance	The object or shielding is likely to transmit the sunlight to your vicinity
	Solar heat transmittance	The object or shielding is likely to transmit the solar heat to your vicinity
	Visibility	I feel that my view is obstructed by that object or barrier
	Sense of weight	It feels like an object or shield with weight (e.g. density)
	Degree of outdoor	How much do you feel separated from the outdoor environment by the object or barrier?
Spatial evaluation	Degree of outdoor	Overall, how separated do you feel the space at this location is from the outdoor environment?
Thermal environment evaluation	Degree of outdoor	How close do you consider the thermal environment of this location to be to the outdoors?

Comparison of the impact of air velocity fluctuations on pleasantness/unpleasantness by surveyed space Real-time voting system

Figure 4 shows the real-time sensation voting system. Using a system developed by Nakagawa et al.¹⁸ for voting on pleasantness/unpleasantness using a smartphone, we surveyed the immediate psychological quantity of subjects. Subjects could vote multiple times in 10 s by hitting the “good” or “bad” button repeatedly, expressing the intensity of pleasantness/unpleasantness by the number of votes. Subjects were instructed to press the button when “feeling thermal comfort or discomfort.”

Variations existed in the number of real-time sensation votes by each subject throughout the day. Thus, for the analysis below, we used the Real-time Pleasantness Vote Index (RPV) calculated with Equations (1) and (2). RPV is a numerical value obtained by normalizing the 30-s running average for the difference between the number of good votes and bad votes (gb) using the mean and standard deviation of gb for each subject so that the mean and standard deviation would be 0 and 1, respectively. It is a relative index where an average state of pleasantness/unpleasantness for each subject was used as the reference. Therefore, RPV of >0 implies thermal comfort that is higher than the mean, and RPV of <0 implies thermal comfort that is lower than the mean.[Image Omitted. See PDF] [Image Omitted. See PDF]where G(t) is number of good votes in the past 10 s at time t, B(t) is number of bad votes in the past 10 s at time t, RPV is real-time pleasantness vote index, gb is 30-s running average for the difference between good and bad votes, gb is mean gb (for each subject), σ_gb is standard deviation of gb (for each subject), and t indicates is time [seconds].

Power regression analysis using a hierarchical Bayesian model

Figure 5 shows the statistical model for RPV prediction in this study. As a statistical analysis method, we used a Hierarchical Bayesian model (HBM). As the approximation equation, we employed a “power regression equation” reflecting the exponential increase and decrease in the target variable relative to the explanatory variable. HBM is a method that creates a hierarchical model incorporating individual and group differences of data and obtaining the posterior distribution of regression parameters that consider uncertainties by using the Bayes' theorem. By using HBM, Lim et al.^23–25 proposed a model to predict the percentage of dissatisfaction considering the variations due to individual differences in thermal sensation. As such, Bayesian estimation is effective in regression analysis of data with variations. A method of Bayesian estimation, HBM, is well-suited to analysis that considers individuality, such as individual differences and differences in surveyed spaces. We estimated regression parameters with the Markov chain Monte Carlo (MCMC) method by using a statistical model of (3), which incorporates the individual difference that the level of impact air velocity has on RPV for each surveyed space by assuming that RPV follows a normal distribution.[Image Omitted. See PDF]where RPV is real-time pleasantness vote index, Va 30-s running average for air velocity [m/s], σ is error, a[i], b[i], and c[i] are regression parameters for surveyed space i, i is surveyed space number, and a₀, b₀, and c₀ are regression parameters (individual difference part for surveyed space). For Bayesian estimation of regression parameters, we used Stan,²⁶ which is a probabilistic programming language for statistical inference. Table 3 shows the setting conditions for MCMC sampling. The number of sampling iterations was 23 000. The first 3000 iterations were the warmup section for the estimate.

Table 3 Setting conditions for MCMC sampling

	Set value
Iteration [−]	23 000
Warmup [−]	3000
Chains [−]	3
Seed [−]	123
Upper and lower parameter limits [−]	−5 < a < 5
	−5 < b < 5
	0 < c < 10

Result of Physical Environment Measurements and Grade Classification Based on Impression Evaluation Thermal environment in semi-outdoor spaces

Table 4 shows the weather, physical environment measurements for the fixed outdoor points, and subject information. The mean air temperature was 22.8°C–29.9°C, the mean relative humidity was 42.3%–85.1%, and the MRT was 26.0 °C–41.7°C. The maximum air velocity was 5.50 m/s or less for each day. In this study, measurement days with the highest temperature of 30°C or more were defined as “summer condition,” while measurement days with the maximum temperature below 30°C were defined as “intermediate condition” for comparison.

Table 4 Weather and thermal environment at outdoor measurement point and subject information

		9/9	9/10	9/13	9/14	9/15	9/17	9/25
Climatic condition		Intermediate	Summer	Summer	Intermediate	Summer	Intermediate	Intermediate
Weather		Cloudy	Sunny	Sunny	Cloudy	Cloudy	Cloudy	Cloudy
Air temperature [°C]	Max.	23.8	33.8	33.3	26.2	31.1	26.0	24.5
	Min.	22.3	27.7	27.3	24.8	23.9	22.5	22.7
	Avg.	22.8	29.6	29.9	25.5	27.6	24.6	23.4
Relative humidity [%]	Max.	87.3	63.2	64.0	69.1	69.0	66.8	68.7
	Min.	81.0	42.0	25.5	58.4	43.5	57.3	55.8
	Avg.	85.1	53.5	42.3	64.1	54.5	62.7	63.6
Air velocity [m/s]	Max.	1.86	5.50	3.50	3.85	3.92	3.94	4.16
	Min.	0.00	0.00	0.24	0.00	0.15	0.12	0.18
	Avg.	0.62	1.12	1.22	0.94	1.10	1.01	1.08
MRT [°C]	Max.	30.1	62.1	55.7	35.0	57.7	41.0	37.4
	Min.	24.2	27.1	28.2	25.7	28.0	25.6	22.4
	Avg.	26.0	41.1	41.7	28.9	41.7	30.6	26.6
Number of subjects (male: female) [−]		4 (2:2)	6 (3:3)	5 (4:1)	3 (2:1)	6 (3:3)	3 (2:1)	7 (4:3)
Clothing insulation [clo]	Avg.	0.82	0.59	0.61	0.63	0.71	0.55	0.69
	SD	0.14	0.08	0.08	0.12	0.12	0.02	0.14

Figure 6 shows the measurement of the physical environment (air temperature, air velocity, and operative temperature) for surveyed spaces along with SET*. SET* was calculated using the estimated values of clothing insulation for each subject and metabolic rate of 1.0 met. The difference in mean between surveyed spaces under the summer condition was the maximums of 2.8 K for the air temperature, 3.8 K for the operative temperature, and 0.77 m/s for the air velocity, showing differences by space. The summer mean of SET* was the highest (31.5°C) for the student lounge and the lowest (26.4°C) for the terrace. Similarly, the differences in the mean by surveyed space under the intermediate condition were the maximums of 2.0 K for air temperature, 3.5 K for operative temperature, and 0.36 m/s for air velocity, showing the differences between spaces. The mean SET* under the intermediate condition was the highest (28.6°C) for the student lounge and the lowest (24.0°C) for the terrace. The mean operative temperature of four spaces excluding the student lounge was 28.0 to 30.0°C for the summer condition and 24.5 to 26.5°C for the intermediate condition, and a notable difference in the operative temperature distribution for each surveyed space has not been presented. The student lounge had a large ratio of FIX window to opening, reducing the air velocity. Simultaneously, it was not air-conditioned; thus, the mean operative temperature was higher than that of the other spaces.

Impression evaluation of surveyed spaces “Degree of outdoor” evaluation of spatial impression and thermal environment

Figure 7 shows the result of “degree of outdoorness” [Note 6] evaluation for the thermal environment along with the impression of the spaces. The order of each surveyed space with the highest degree of outdoorness was “courtyard > terrace > piloti > atrium > student lounge.” Subjects considered surveyed spaces with a higher degree of outdoorness as being unsteady and nonuniform environments.

Classification of environmental grade

In this study, we focus on the top surface (ceiling and trees, etc.) and elevation (walls, opening, surrounding buildings, and so on) that constitute the spaces and propose a method that classifies semi-outdoor spaces into two environmental grades. As shown in Figure 8-1, we plotted the pre-rooms as the origin and values of the fixed outdoor points as the maximum values on a two-dimensional plane, ordered by distance from the origin. As for the physical environment, we used short wavelength radiation and air velocity that likely reflect the degree of openness for the top surface and elevations, and plotted each surveyed space. The result in the order of outdoor tendency was “courtyard > terrace > piloti > student lounge > atrium.” Distribution of surveyed spaces using the physical environment did not present a clear difference in the plot location of surveyed spaces excluding the courtyard; thus, environmental grade classification was not possible. Next, for the impression evaluation, we used the evaluation of “degree of outdoorness” for the top surface and elevations (weighted mean of four surfaces [Note 7]) and plotted each surveyed space on a two-dimensional plane, as shown in Figure 8-2). The order of each surveyed space in terms of outdoor tendency was “courtyard > terrace > atrium > piloti > student lounge.” Evaluation for the elevations only was “courtyard > terrace > piloti > atrium > student lounge,” and the evaluation for the top surface only was “terrace > courtyard > atrium > student lounge > piloti.” On the two-dimensional plane plot (see Figure 8-2)) that showed the difference in the order of “outdoorness” between the top surface and elevations, plot positions for “courtyard and terrace” and “atrium, piloti, and student lounge” were separated; thus, we classified them into two environmental grades: “outdoor tendency grade” and “indoor tendency grade.”

Influence structure of each question item on “degree of outdoor” evaluation

To clarify the structure of the impact of each question (see Table 2) on “the degree of outdoorness” for the elevations and the top surface, we performed a principal component analysis and multiple regression analysis. We derived the partial regression coefficient for each principal component against the voting value of the evaluation for “the degree of outdoorness” for the elevations and the top surface, and compared the degree of impact for each element.

Figure 9 shows the relationship between “the degree of outdoorness” for the elevations and each question. For the first principal component (PC1), the absolute value of eigenvector for each question was 0.45–0.53: a variable that showed “comprehensive openness.” The second (PC2) and the third (PC3) principal components were variables with a large impact of “visibility shielding” and “breeziness,” respectively. The absolute value of the partial regression coefficient for PC1 and PC2, which is the evaluation of “the degree of outdoorness” for elevation, was the same and about 0.17 higher than that of PC3. Therefore, we found that following “comprehensive openness” and “visibility shielding” notably impacted the evaluation of “the degree of outdoorness.”

Figure 10 shows the relationship between “the degree of outdoorness” for the top surface and each question. Similar to the elevations, for PC1, the absolute value of eigenvector for each question was 0.45–0.55: a variable that shows “comprehensive openness.” PC2 and PC3 were variables with a high degree of impact of “sense of weight” and “visibility shielding.” The partial regression coefficient for the evaluation of “the degree of outdoorness” for the top surface was 0.13 higher for PC1 than PC2, where PC3 was close to zero. Thus, following “comprehensive openness,” “sense of weight” notably impacted “the degree of outdoorness.”

Psychological quantity

Figure 11 shows the result of each psychological quantity voting (thermal sensation, thermal comfort, thermal preference, and air movement sensation) for each surveyed space. A total of 30 men and women answered the questions regarding thermal sensation, thermal comfort, and thermal preference for five spaces and voted 11 times for each surveyed space. Thus, the total number of samples reached 330. Air movement sensation and comfort with air movement were only voted on once for each surveyed space; thus, the number of samples for each surveyed space was 30.

Figure 11-1) shows the result of thermal sensation voting for each surveyed space. Under the summer condition, many voted “cool” for the terrace, atrium, and courtyard, while relatively higher votes were cast for “warm” for the student lounge and piloti. Under the intermediate condition, a relatively large number of votes said “cool” for all spaces excluding the student lounge.

Figure 11-2) shows the result of the thermal comfort voting for each surveyed space. The overall trend was more votes for comfort than discomfort [Note 8]). Under the summer condition, a relatively larger number of votes were cast for “comfortable” for the terrace, atrium, and courtyard, while a relatively larger number of votes were cast for “uncomfortable” for the student lounge. Under the intermediate condition, a relatively larger number of votes were cast for “comfortable” for all surveyed spaces.

Figure 11-3) shows the result of the thermal preference voting for each surveyed space. Under the summer condition, there were more preference votes for “cooler” in all surveyed spaces [Note 9]. At the student lounge where many voted for discomfort under the summer condition, with two-thirds of votes for “cooler.” Since SET* was higher than in other surveyed spaces and fluctuations in the air velocity were limited, heat was more noticeable. Contrarily, under the intermediate condition, relatively more answers said that “as it is” for all surveyed spaces. Since under the summer condition, the terrace and courtyard had many comfortable votes and the mean thermal preference vote was close to 0, which implied that “as it is,” and subjects were mostly satisfied with the environment. The mean air movement sensation vote for the terrace and courtyard was 2 or higher, which implies “feeling”; thus, it is possible that the presence of adequate airflow led to improvements in thermal comfort.

Figure 11-4) shows the result of the air movement sensation voting for each surveyed space. Under the summer and intermediate conditions, the mean vote for air movement sensation was high in the order of “courtyard > terrace > atrium > piloti > student lounge” and “courtyard > terrace > piloti > atrium > student lounge,” respectively. Excluding the atrium in the summer, the results mostly corresponded with the size of mean air velocity shown in Figure 6-2).

Summary

We showed measured physical environment values and SET* for each surveyed space and plotted the measured values of downward short wavelength radiation and air velocity along with the evaluation of “the degree of outdoorness” for the top surface and elevations, aiming to order each surveyed space. Among the impression evaluation, by using the evaluation of “the degree of outdoorness” for the top surface and the elevations, we classified surveyed spaces into “indoor tendency grade” and “outdoor tendency grade.” Based on the principal component analysis, we showed the impact of each question on the evaluation of “the degree of outdoorness.” Additionally, we compared surveyed spaces for the voting of psychological quantities. In the next section, we use the relationship of SET* and the percentage of uncomfortable vote for each surveyed space to order spaces. We will discuss the corresponding relationship with environmental grade classification based on the impression evaluation.

Discussion on Thermal Comfort Effect of differences in climatic conditions on thermal comfort

Figure 12 [Note 10] shows the relationship between SET* and the percentage of uncomfortable vote for each climatic condition. SET* was rounded up to integers (unit of 1°C). Discomfort in terms of the thermal comfort and requesting “cooler” or “warmer” for the thermal preference (excluding voting for “no change”) were considered votes of dissatisfaction. We also performed probit regression [Note 11] and compared the predicted percentage of dissatisfied⁶ (PPD) calculated for the standard condition of SET* (air temperature = radiation temperature, air velocity = 0.1 m/s, amount of clothing = 0.6 clo, and metabolic rate = 1.0 met). Under the intermediate condition, the percentage of uncomfortably cool votes in the range of SET* (24.0°C–27.0°C) was higher than PPD.

Between the summer and intermediate conditions, requests in terms of thermal preference were different. As shown in Figure 13, 84% of the dissatisfied reports under the intermediate condition requested “warmer.” Nikolopoulou et al.²⁷ showed that the neutral temperature changes by seasons, being higher in the summer and fall compared to the winter. The mean high temperature for Tokyo in August was 31.6°C,²⁸ although the mean high temperature for the measurement days under the intermediate condition was 25.4°C. Therefore, subjects for the present measurements were adapted to the summer climate and felt the measurement days under the intermediate condition were cool, leading to an increase in uncomfortably cool votes for the range of SET* 24.0°C–27.0°C.

Relationship between SET* and the percentage of uncomfortable votes at each surveyed space

Figure 14 shows the relationship between SET* and the percentage of uncomfortable vote for each surveyed space. In this study, we defined the SET* that leads to the rate of dissatisfied report being less than 20% according to ASHRAE Standard 55¹⁰ as the thermal comfort zone. For the terrace and courtyard, the percentage of uncomfortably warm votes was constantly less than 20%. Specifically, the percentage of uncomfortable vote for the terrace was the lowest among all surveyed spaces. The reason for this may be that the range of SET* obtained at the time of reporting for the terrace was slightly narrower than other surveyed spaces, and 74% of the dissatisfied reports occurred within the SET* range of 25.0°C–27.0°C.

According to the regression curve of uncomfortably warm votes, SET* showing the upper limit of the thermal comfort zone was 30.2°C, 28.2°C, and 28.4°C for the atrium, student lounge, and piloti, respectively. Compared to the upper limit of PPD, that is, 27.8°C, each upper limit was higher, and the thermal comfort zone for the semi-outdoor environment expanded compared to the indoor spaces. Among those, we compared the courtyard that is classified to the outdoor tendency grade and the piloti that is classified to the indoor tendency grade according to the impression evaluation. The result showed that the distribution of SET* was 22.0 to 32.0°C for both, as shown in Figure 6-4). However, if the regression curve was extrapolated to SET* of 33.0°C, the upper limit of the thermal comfort zone would likely be higher for the courtyard than the piloti. The reason for this is that the difference in psychological adaptation caused by the different expectation for the environment led to expansion of the upper limit of the thermal comfort zone.

With the regression curve of uncomfortably cool votes, SET* at the lowest limit for the thermal comfort zone was 26.7°C, 23.6°C, and 25.5°C for the atrium, courtyard, and piloti, respectively. Compared to the lower limit for PPD (22.9°C), each lower limit was higher.

Relationship between environmental grade classification based on impression evaluation and environmental control methods

Figure 15 shows the percentage of uncomfortably warm votes separated for outdoor and indoor tendency grades. Based on the result of the previous section, to compare the difference in environmental grades and thermal comfort, we focused on the percentage of uncomfortably vote requesting “cooler,” and compared the range of increase (ΔUW) for the percentage of uncomfortably warm votes in the range of SET* (25.0–30.0°C) for the result of the present study and the measurements taken for the air-conditioned atrium (HVAC space) and non-air-conditioned galleria and wood deck (non-HVAC space) by Nakano et al.⁷ The courtyard and terrace which are classified into outdoor tendency grade had low ΔUW similar to the non-HVAC space, while the piloti, atrium, and student lounge which are classified into indoor tendency grade had high ΔUW similar to the HVAC space. Thus, it was shown that the grade classifications based on the environment control method of Nakano et al. and on the impression evaluation in this study correspond in terms of thermal comfort.

Discussion on evaluation methods for environmental grade classification

Table 5 shows the relationship between grade classifications based on the impression evaluation and thermal comfort. For three surveyed spaces of indoor tendency grade, where the upper limit of the thermal comfort zone, as shown in Figure 15-2) was obtained, we compared the result of each evaluation question with ΔUW. The order of surveyed spaces based on the physical environment and the order of ΔUW were not consistent; and thus, evaluating the thermal comfort for the semi-outdoor environment based on physical environment alone would be difficult. Next, the order of the evaluation of “the degree of outdoorness” obtained by plotting the mean votes for the top surface and the elevations on a two-dimensional plane was consistent for the atrium with the lowest ΔUW of the indoor tendency grade, where it had an impression evaluation relatively close to the outdoor tendency. However, the orders of piloti/student lounge and terrace/courtyard were switched. The order of surveyed spaces based on the evaluation of “the degree of outdoorness” for the top surface only was consistent with the order of ΔUW. According to the impression evaluation of the previous section, especially when “sense of weight” decreased, voting for the evaluation of “the degree of outdoorness” for the top surface tended to increase. For the atrium that uses glass for the ceiling window, roof, and terrace, and the courtyard that is shaded by trees, subjects did not feel much of the “sense of weight” of the top surface, leading to an evaluation toward outdoor tendency. Thus, the expectation for the thermal environment was eased and the thermal comfort zone expanded compared to the PPD. This showed the possibility of evaluating the thermal comfort of diverse semi-outdoor environment with spatial characteristics in addition to differences in the environmental control.

Table 5 Relation between grading based on impression evaluation and thermal comfort

Evaluation method for grading		Outdoor ←				→ Indoor
Physical environment	Downward short wavelength radiation and air velocity	C (Courtyard)	A (Terrace)	E (Piloti)	D (Student lounge)	B (Atrium)
Impression evaluation	Degree of outdoor (thermal environment evaluation and spatial evaluation)	Outdoor tendency		Indoor tendency
		C	A	E	B	D
	Top surface and elevation	C	A	B	E	D
	Elevation	C	A	E	B	D
	Top surface	A	C	B	D	E
Increase in percentage of uncomfortably warm votes in the range of SET*25.0°C to 30.0°C (ΔUW)		Small ←				→ Large
Previous studies by Nakano et al.7		Non-HVAC space ΔUW = 9.5		HVAC space ΔUW = 20.5
This study		A (Terrace)	C (Courtyard)	B (Atrium)	D (Student lounge)	E (Piloti)
		ΔUW = 3.8	ΔUW = 8.3	ΔUW = 16.8	ΔUW = 27.1	ΔUW = 45.6

Summary

In this section, we showed the relationship between SET* and percentage of uncomfortable vote. By deriving the approximation curve through probit regression, and obtained the upper and lower limits for the thermal comfort zone for each surveyed space. Comparison of the courtyard (outdoor tendency grade) and the piloti (indoor tendency grade) with a common distribution of SET* showed that the upper limit of SET* for the thermal comfort zone would be higher for the courtyard. Therefore, the difference in psychological adaptation such as easing of expectation toward environment impacted the expansion of the thermal comfort zone. In terms of the regression curve for the percentage of uncomfortably warm votes, we compared the environmental grade classification based on the existing studies and impression evaluations. In addition to the difference in the environmental control, the difference in spatial characteristics may have impacted the expansion of the thermal comfort zone in the semi-outdoor environment. In the next section, we present the use of a real-time sensation voting survey and HBM to clarify the impact of environmental grade on the correlation between air velocity and pleasantness/unpleasantness of occupants.

Derivation of Regression Equation for Real-Time Pleasantness Vote Index on Air Velocity and Comparison by Surveyed Space Real-time pleasantness vote index at each surveyed space

Figure 16 shows the RPV distribution for each surveyed space. The order of mean RPV under the summer condition was “courtyard > terrace > atrium > student lounge > piloti.” As shown in Figure 6-2), the mean RPV for the atrium, student lounge, and piloti with mean air velocity of less than 0.5 m/s was less than 0. The mean RPV for the terrace and courtyard with the mean air velocity of 0.5 m/s or higher was higher than 0. The order of mean RPV under the intermediate condition was “terrace > student lounge > piloti > courtyard > atrium.” Its trend was different from that of the summer condition. Under the intermediate condition, the mean RPV was higher than 0 for the terrace and student lounge, but less than 0 for other surveyed spaces. Under the intermediate condition, the student lounge had the highest operative temperature among all surveyed spaces, leading to an increase in RPV. Air velocity and operative temperature for the terrace were the same as values measured at other surveyed spaces except for the student lounge. Thus, the elements other than thermal physical environment (impression on light, sound, and space, etc.) likely impacted the increase in RPV.

Correlation between real-time pleasantness vote index and psychological quantity

Figure 17 shows the correlation between RPV and psychological quantity voting (thermal sensation, thermal comfort, thermal preference, and air movement sensation). RPV was rounded to multiples of 0.5. The mean of psychological quantity voting belonging to each RPV category was calculated and graphed. The size of the bubble refers to the number of responses. We performed linear regression by the least-square method and obtained the primary approximation equation that expresses the correlation between RPV and each voting for psychological quantity.

Correlation with thermal sensation vote

Figure 17-1) shows the correlation between RPV and thermal sensation. Under the summer condition, the slope of the regression line of RPV relative to the thermal sensation was −0.36, and 0.15 under the intermediate condition. Thus, the thermal comfort improved when subjects sensed coolness or warmth under the summer or intermediate conditions, respectively, which led to increased RPV.

Correlation with thermal comfort vote

Figure 17-2) shows the correlation between RPV and thermal comfort. Under the summer condition, the slope of the regression line for RPV related to the thermal comfort was 0.59. It was 0.33 under the intermediate condition. Correlation of thermal comfort and RPV under the summer condition was stronger than that under the intermediate condition [Note 12]. Additionally, under both conditions, there was a positive correlation between thermal comfort and RPV; thus, it was confirmed that the number of real-time sensation votes by subjects expressed the strength of pleasantness/unpleasantness.

Correlation with thermal preference vote

Figure 17-3) shows the correlation between RPV and thermal preference. Under the summer condition, the slope of the regression line of RPV relative to thermal preference voting was −0.07, while it was 0.06 under the intermediate condition. Through both conditions, when subjects felt the thermal sensation was comfortable, RPV increased.

Correlation with thermal preference vote

Figure 17-4) shows the correlation between RPV and air movement sensation [Note 13]. Under the summer condition, the slope of the regression line for RPV relative to air movement sensation voting was 0.43. It was 0.13 under the intermediate condition. The correlation between air movement sensation and RPV under the summer condition was stronger than that under the intermediate condition. More airflow being sensed led to an improvement in the thermal comfort.

Regression analysis with HBM

Air velocity distribution varies by surveyed space (Figure 6). Correlation between RPV and psychological quantity differed between the summer and intermediate conditions (Figure 17). Therefore, we separated the measurement data of the terrace, courtyard, and piloti with a wide distribution of air velocity, and the atrium and student lounge with a narrow distribution, and conducted a regression analysis with HBM for each climatic condition. Although there were individual differences among subjects, the overall trend of RPV in this study was close to a normal distribution.

Estimation results for regression parameters

Table 6 shows the estimate of power regression parameters. We used the mean of the posterior distribution for each parameter as the regression coefficient. To determine the convergence in the MCMC method, an index called Rhat was used. Generally, Rhat ≤1.1 is considered “converged.” In this analysis, we saw Rhat ≤1.1 for estimates of all parameters, confirming convergence of the MCMC sampling.

Table 6 Estimation results for the parameters of the power regression equations

		Summer condition
		a		b		c
		Mean	SD	Mean	SD	Mean	SD
Wide air velocity distribution	Common part	−0.53	1.05	1.07	1.04	1.74	1.11
	A: Terrace (i = 1)	−0.16	0.22	0.69	0.26	1.73	0.43
	C: Courtyard (i = 2)	−0.65	0.50	1.29	0.51	0.90	0.42
	E: Piloti (i = 3)	−0.82	0.08	1.41	0.13	1.76	0.24
Narrow air velocity distribution	Common part	−0.29	1.77	0.78	2.27	4.37	2.46
	B: Atrium (i = 1)	−0.09	0.05	1.25	2.31	5.59	2.19
	D: Student lounge (i = 2)	−1.00	0.60	2.42	0.71	1.11	0.67
		Intermediate condition
		a		b		c
		Mean	SD	Mean	SD	Mean	SD
Wide air velocity distribution	Common part	−0.07	0.91	0.01	1.16	4.10	1.70
	A: Terrace (i = 1)	0.33	0.11	−0.22	0.17	3.87	1.71
	C: Courtyard (i = 2)	−0.15	0.04	−0.40	0.12	4.84	1.35
	E: Piloti (i = 3)	−0.36	0.05	0.62	0.09	3.35	0.54
Narrow air velocity distribution	Common part	−0.14	1.97	−0.21	2.59	5.37	2.62
	B: Atrium (i = 1)	−0.54	0.16	−0.11	2.58	5.66	2.51
	D: Student lounge (i = 2)	0.22	0.07	−0.41	2.62	5.71	2.43

Relationship between running average of air velocity and real-time pleasantness vote index and regression curve at each surveyed space

1. Survey space with air velocity distribution of 0 m/s–2.0 m/s (summer condition)

Figure 18-1) shows the scatter plot for the running average for air velocity and RPV under the summer condition for the terrace, courtyard, and piloti, and the power regression curve for each surveyed space. The dashed line shows the regression curve that indicates the trend of the overall data (common part), while the solid line is a curve that shows the trend of each surveyed space. The trend of the overall data is expressed with the regression curve, where an increase in air velocity led to an increase in the RPV index.

When we focused on the regression curve of the piloti, the area with air velocity below 1.0 m/s had RPV that was lower than surveyed spaces classified under outdoor tendency grade. The area with air velocity of 1.0 m/s or higher had a large rate of increase for RPV. As shown in Figure 6-3), the mean operative temperature for the terrace, courtyard, and piloti under the summer condition was within the range of 28.0°C–30.0°C, and there was no notable difference in the operative temperature distribution. Therefore, it is possible that factors other than thermal physical environment impacted on the pleasantness/unpleasantness for subjects.

Auliciems²⁹ showed that the “expectation for the environment” may impact the thermal comfort. In this study, we focus on the “expectation for airflow” and discuss the reason for the difference in the correlation between air velocity and RPV between surveyed spaces. Figure 19 shows the distribution of the “breeziness” evaluation for the elevations under the impression evaluation of each surveyed space. The mean “breeziness” evaluation of the elevations was high in the order of “courtyard > terrace > piloti > atrium > student lounge,” while the piloti had a lower value than that of the terrace or courtyard. The fact that subjects considered the piloti as an area that is not breezy and had a low expectation for air flow led to an increase in the RPV index when the air velocity was high (air velocity of 1.0 m/s or higher).

Therefore, under the summer condition, semi-outdoor spaces should be designed while considering the impact of “expectation for airflow” on the thermal comfort. Specifically, for semi-outdoor spaces that are classified under indoor tendency grade and have low expectation for airflow, making a plan that assures airflow pathway considering the main wind direction under the summer condition would be effective.

The regression curve for the terrace and courtyard had similar slope and positional relationship in the scatter plot. Therefore, it is assumed that the impact of air velocity on RPV for both surveyed spaces has a similar trend. The reason for this is that both spaces are of outdoor tendency grade with tree cover where expectation for airflow was high compared to spaces of indoor tendency grade such as the piloti.

1. Survey spaces with the air velocity distribution of 0 m/s–2.0 m/s (intermediate condition)

Figure 18-2) shows the scatter plot of running average for air velocity and RPV for the terrace, courtyard, and piloti under the intermediate condition and the power regression curve for each surveyed space. The regression curve for the common part was mostly horizontal, and the correlation between air velocity and RPV was not confirmed for the trend of the overall data under the current condition.

The intercept “a” for the regression equation of the terrace was 0.33. RPV was higher than other surveyed spaces for the air velocity range of 0 m/s–1.0 m/s; however, the environment, where air velocity was 1.0 m/s or higher, had a tendency to have lower RPV. As shown in Figure 17-1), under the intermediate condition, RPV dropped when thermal sensation voting was low, and an increase in air velocity led to discomfort with cold sensation. This trend was similar for the courtyard where a higher air velocity led to lower RPV.

The intercept “a” for the regression equation of the piloti was −0.36, and RPV was lower than the other survey points in the air velocity range of 0 m/s − 0.75 m/s. As the air velocity increased, RPV index increased. This is a trend different from other surveyed spaces. Since operative temperature distribution of each surveyed space did not present a notable difference, factors other than the thermal physical environment impacted the correlation between air velocity and RPV.

In this study, we were unable to clarify the cause that had a notable impact; thus, additional surveys should be performed for the intermediate period for verification.

1. Survey spaces with the air velocity distribution of 0 m/s–0.6 m/s (summer condition)

Figure 18-3) shows the scatter plot of the running average for air velocity and RPV for the atrium and student lounge under the summer condition and the power regression curve for each surveyed space. The regression curve for the common area was mostly horizontal, and the correlation between air velocity and RPV was not confirmed in the overall data trend under this condition.

The regression curve of the atrium was also almost horizontal, and there was no correlation between air velocity and RPV. For the student lounge, there was mostly positive linear relationship between air velocity and RPV. Figure 19 shows that the student lounge with a large opening with a large relative area of the FIX window had the lowest mean “breeziness” evaluation for the elevations among all surveyed spaces. Therefore, the student lounge had a lower expectation for natural wind compared to the atrium, increasing sensitivity to the airflow. As shown in Figure 6-3), since the mean operative temperature of the student lounge was the highest among all surveyed spaces, the difference in the operative temperature distribution may have had an impact.

1. Survey spaces with air velocity distribution of 0 m/s–0.6 m/s (intermediate condition)

Figure 18-4) shows a scatter plot of the running average for air velocity and RPV for the atrium and student lounge in the intermediate period and the power regression curve of each surveyed space. The regression curve of the common part is mostly horizontal, and correlation between air velocity and RPV was not confirmed in the trend in the overall data or each surveyed space in this condition.

Summary

In this section, we showed the correlation between RPV and psychological quantity voting and confirmed the relationship between thermal comfort and real-time sensation voting of subjects. Additionally, using HBM, we obtained the power regression equation of RPV relative to the running average for air velocity for each surveyed space. If we focus on the analytical result of the summer condition, the running average for air velocity and RPV had a correlation that should be expressed with a positive power approximation curve for surveyed spaces with air velocity distribution of 0 m/s–2.0 m/s.

The regression curve of the piloti had a larger increase of RPV compared to that of the other surveyed spaces for air velocity of 1.0 m/s or higher, although there was no notable difference in the operative temperature distribution for the terrace, courtyard, and piloti. Since factors other than the thermal physical environment impacted the correlation between air velocity and RPV, we focused on the “expectation for airflow” and discussed the result of the impression evaluation of spaces. Based on the result of “breeziness” evaluation of the elevations, we found that subjects considered the piloti to be not breezy and had little expectation for airflow, which led to an increase in the RPV index for air velocity range of 1.0 m/s or higher.

Previous studies exist in which subject experiments that simulated natural wind with an artificial climate chamber showed the relationship between airflow and thermal comfort^30,31; however, hardly any studies are available that clarified the relationship between air velocity and thermal comfort in the semi-outdoor environment while considering the difference in “expectations for the environment” in spatial characteristics. In this study, we combined the tendencies in the “breeziness” evaluation of the elevations with the correlation between air velocity and the pleasantness/unpleasantness for analysis and showed that the difference in the “expectations for airflow” in spatial characteristics impacted the correlation between air velocity and thermal comfort.

Conclusion

With the aim of understanding the impact of different environmental grade on thermal comfort of occupants in semi-outdoor spaces, we conducted physical environment measurements, questionnaire, and a real-time sensation voting survey at five spaces at Nishiwaseda Campus of Waseda University, and obtained the following findings:

1. We proposed a method to classify the semi-outdoor environment into two environmental grades based on the impression evaluation of spaces. Using the “degree of outdoorness” evaluation for the top surface and the elevations, we plotted each surveyed space on a two-dimensional plane. Plot positions for “courtyard and terrace” and “atrium, piloti, and student lounge” were separated; thus, these spaces were classified into outdoor and indoor tendency grades: two environmental grades.
2. The principal component analysis showed that the principal component showing “comprehensive openness” for the elevations and top surface mostly impacted the evaluation of “the degree of outdoorness.” For the evaluation of “the degree of outdoorness” of the elevations, the principal component that shows “visibility shielding” had the highest impact following “comprehensive openness.” For the evaluation of “the degree of outdoorness” for the top surface, the principal component that shows “sense of weight” had the highest impact following “comprehensive openness.”
3. Under the summer condition, comfort voting was high for the terrace and courtyard and the mean thermal preference vote was close to 0, which indicates “as it is”; thus, subjects were mostly satisfied with the environment. The mean air movement sensation vote for the terrace and courtyard was two or higher, which indicates “feeling”; thus, presence of adequate airflow lead to an increase in thermal comfort.
4. For ΔUW, for the percentage of uncomfortably warm votes for the SET* range of 25.0°C–30.0°C, we compared the trend for surveyed spaces in this study with existing studies. In the impression evaluation, surveyed spaces that are classified into the outdoor tendency grade (terrace and courtyard) had low ΔUW similar to that in the non-HVAC spaces in earlier studies, while surveyed spaces that are classified into the indoor tendency grade (atrium, student lounge, and piloti) had high ΔUW similar to that in the HVAC spaces. Thus, the classification of surveyed spaces based on the different environment control methods and the grade classification based on the impression evaluation for spaces correspond in terms of thermal comfort.
5. The order of “the degree of outdoorness” evaluation for the top surface for each surveyed space was consistent with the order of ΔUW. The atrium with a glass ceiling window and the terrace and courtyard shaded by trees had reduced “sense of weight” of the top surface, which eased the expectations for the thermal environment, which in turn expanded the thermal comfort zone. The difference in thermal comfort for the diverse semi-outdoor environment could be evaluated by the difference in spatial characteristics such as “sense of weight” of the top surface in addition to the difference in environment control.
6. With the real-time sensation voting system, we acquired the information on pleasantness/unpleasantness experienced by subjects instantly. With this result, we performed a power regression analysis that considered individual differences in regression parameters for surveyed spaces using HBM. In the semi-outdoor environment with air velocity distribution of 0 m/s–2.0 m/s, there was a correlation between the running average for air velocity and RPV that can be expressed with a power approximation curve. The regression curve of the piloti had a higher range of increase in RPV compared to that of other surveyed spaces for air velocity of 1.0 m/s or higher. However, there was no noticeable difference in the operative temperature distribution for the terrace, courtyard, and piloti. Factors other than thermal physical environment may have impacted the correlation between air velocity and RPV. Thus, we focused on the “expectation for airflow” and discussed the result of the impression evaluation for spaces. The result of the “breeziness” evaluation for the elevations showed that subjects considering the piloti as not breezy and having low expectation for air flow led to RPV index increasing for the air velocity range of 1.0 m/s or higher.
7. Under the intermediate condition, the regression curve of the piloti had a notable difference from the tendency of the terrace and courtyard; thus, the factors other than thermal physical environment impacted the correlation between air velocity and RPV. In this study, we could not clarify the specific factor that impacted and the extent of impact; thus, we need additional surveys for the intermediate period for verification.

Going forward, we need to conduct surveys in more diverse semi-outdoor spaces and discuss environmental grade classification and thermal comfort. Further experiments should be conducted in the future to verify the validity and effectiveness of RPV. This article was prepared by making additions and corrections to the earlier reports.^32–34

Acknowledgments

This study was funded and subsidized by Grants-in-Aid for Scientific Research (Grant-in-Aid for Scientific Research (A): Project No. 19H00797, Grant-in-Aid for JSPS Fellows: Project No. 20J14702, Grant-in-Aid for Research Activity Start-up: Project No. 20K22449). Part of this study was conducted with the support of Obayashi Foundation. It is part of a project for Waseda Research Institute for Science and Engineering, Waseda University.

Disclosure

The authors have no conflict of interest to declare.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

NOTES:

^{Note 1)}This measurement was approved by the ethics commit of Waseda University (2021–102).

^{Note 2)}The number of subjects for each measurement day is shown in Table 4. In principle, different subjects participated each day. On September 9, the weather was not favorable in the morning; thus, the measurement was only taken in the afternoon. Among subjects who participated on September 9, one male and one female also participated on September 10, and one male participated on September 17, and one female participated on September 25.

^{Note 3)}A mobile measurement cart was placed near subjects or at the center of sitting positions as much as possible.

^{Note 4)}Equations (4)–(7) show calculation formulas for the MRT based on the actual measurement of longwave and shortwave radiations. At fixed outdoor points, we used the equation of Reindl²⁰ to calculate direct and diffuse solar radiations separately from the downward short wavelength radiation.[Image Omitted. See PDF] [Image Omitted. See PDF] [Image Omitted. See PDF] [Image Omitted. See PDF]

^{Note 5)}Uchimura et al.²¹ showed that 9-point thermal sensation scale (“cold – slightly cold – cool – slightly cool – neither – slightly warm – warm – slightly hot – hot”) has an even interval. Thus, the range of −2 to +2 used for the thermal sensation voting in the present study should also have similar intervals. However, an even interval is not confirmed for the areas outside of the thermal sensation voting (−2 to +2) and questions other than thermal sensation. In the present study, we handled all psychological quantity voting as having an even interval for convenience and performed the following analyses in order to compare survey spaces and clarify the correlation with the real-time sensation voting.

^{Note 6)}As shown in Table 2, subjects voted on “the degree of outdoorness” of thermal environment for each elevation surface, top surface, and comprehensive space impression.

^{Note 7)}We used the “degree of outdoorness” evaluation for the front, right, left, and back surfaces as the explanatory variable, and “the degree of outdoorness” evaluation based on the comprehensive space impression as the objective variable, and performed multiple regression analysis. The results showed that the partial regression coefficient of each explanatory variable was 0.42, 0.15, 0.18, and 0.09 for the front, right, left, and back surface, respectively. We weighted “the degree of outdoorness” of each surface with the partial regression coefficient and derived the weighted mean of four surfaces.

^{Note 8)}“Very comfortable,” “comfortable,” and “slightly comfortable” were considered to be “comfortable” votes, and “very uncomfortable,” “uncomfortable,” and “slightly uncomfortable” were considered “uncomfortable” voting.

^{Note 9)}“Cooler” and “warmer” were considered as uncomfortably warm and cool votes, respectively.

^{Note 10)}“Uncomfortably cool” in Figure 12 is the rate of dissatisfied reports requesting warming, and “uncomfortably warm” is the rate of dissatisfied reports requesting cooling.

^{Note 11)}The probit regression equation is shown in Equation (8) (Cumulative distribution function of the standard normal distribution).[Image Omitted. See PDF]

^{Note 12)}RPV is a value normalized by using the mean and standard deviation of gb for each subject (see section 2.4.1); and thus, it has a range of −4 to +4. Therefore, under the intermediate condition where comfortable voting is relatively high, the mean thermal comfort voting was higher than 0 for both situations where RPV >0 and RPV <0.

^{Note 13)}While thermal sensation, thermal comfort, and thermal preference were voted a total of 11 times for each survey space, the air movement sensation was only voted once after staying in a space for 30 min (see section 2.3). Therefore, the number of responses for the air movement sensation was smaller than other questions. The size of each bubble in Figure 17-4) is small compared to Figure 17-1) to Figure 17-3).