Quantitative Study of Using Piloti for Passive Climate Adaptability in a Hot-Summer and Cold-Winter City in China

There has been an insufficient study of passive climate adaptability that considers both the summer and winter season for the outdoor thermal environment of hot-summer and cold-winter cities. In this study, we performed a quantitative simulation to research the passive climate adaptability of a residential area, considering piloti as the main method for climate adaptation in a hot-summer and cold-winter city in China. Numerical simulations were performed with a coupled simulation method of convection, radiation, and conduction. A cubic non-linear k–ε model proposed by Craft et al. was selected as the turbulence model and three-dimensional multi-reflections of shortwave and longwave radiations were considered in the radiation simulation. Through the simulation, we found that setting the piloti at the two ends of the building was the optimal piloti arrangement for climate adaptation. Then the relationship between the piloti ratio (0%, 20%, 40%, 60%, and 80%) and the outdoor thermal environment was studied. It could be concluded that with the increasing piloti ratio, the wind velocity increased, the mean radiant temperature (MRT) decreased slightly, and the average standard effective temperature (SET*) decreased to 3.6 °C in summer, while in winter, with the increasing piloti ratio, the wind velocity, MRT, and SET* changed slightly. The wind environment significantly affected the SET* value, and the piloti ratio should be between 12% and 38% to avoid wind-induced discomfort.


Introduction
With the aggravation of global warming and the expansion of urbanization, serious environmental problems such as air pollution, large anthropogenic heat release, and urban heat island are severer day by day [1][2][3][4]. The Urban Heat Island has caused outdoor thermal environments in residential areas to become worse [5]. Body discomfort and heat stroke rates have increased significantly because of the deterioration of the indoor and outdoor thermal environment [6,7]. Since people feel hotter in the summer, air conditioning is used more frequently and for a longer time; then, the exhaust gas from air conditioning makes the outdoor thermal environment in residential areas worse [8,9]. Therefore, avoiding the vicious cycle of the deterioration of the thermal environment, energy consumption, and further deterioration of the thermal environment' has become an urgent problem in urban development.
China has rapidly increasing energy needs; building energy consumption accounts for approximately one-third of the total social consumption, and this energy consumption is expected people in Wuhan are under conditions of thermal discomfort. The highest temperature could be almost 39 °C, and the lowest temperature could be less than −3 °C.

Prediction Method for the Outdoor Thermal Environment
A coupled approach of convection, radiation, and conduction for the three-dimensional outdoor thermal environment was first initiated by Yoshida et al. based on a self-edited program [34]. They compared the results of simulations with field measurements from a northern part of Tokyo. They used two domains, a large one and a small one; moreover, the large one contained the small one. It was a two-stage coupled analysis. Firstly, a computational fluid dynamics (CFD) analysis was performed in the large area, and measurement values were used as boundary values. They selected a k-ε revised turbulence model and obtained small area's boundary values belonging to the large area. They conducted the CFD analysis in the small area using outcomes in the first step and air temperature and humidity from on-site observations. Finally, they found that results from the numerical analysis agreed with the on-site observations.
For this study, we conducted a non-isothermal CFD analysis using data coming from the local meteorological bureau containing wind velocity, prevailing wind direction, and air temperature, and observed data containing ground and surface temperatures. The ground and building surface temperatures were calculated in step 2 based on the unsteady state heat balance calculation, including three-dimensional radiation and one-dimensional conduction calculations. Based on the results in these two steps, we once again conducted a CFD analysis to obtain more accurate results. Finally, we calculated SET*, a comprehensive index, to evaluate outdoor thermal comfort in the target area, which was in the centre of the analysis domain. Xi [31], Xuan et al. [32] and Zhou et al. [34] conducted their research using a similar prediction method.

Simulation Date
In this study, a weather database of China [35] was used. The typical meteorological day in summer is 1 July and the typical meteorological day in winter is 2 January (Table 1).

Prediction Method for the Outdoor Thermal Environment
A coupled approach of convection, radiation, and conduction for the three-dimensional outdoor thermal environment was first initiated by Yoshida et al. based on a self-edited program [34]. They compared the results of simulations with field measurements from a northern part of Tokyo. They used two domains, a large one and a small one; moreover, the large one contained the small one. It was a two-stage coupled analysis. Firstly, a computational fluid dynamics (CFD) analysis was performed in the large area, and measurement values were used as boundary values. They selected a k-ε revised turbulence model and obtained small area's boundary values belonging to the large area. They conducted the CFD analysis in the small area using outcomes in the first step and air temperature and humidity from on-site observations. Finally, they found that results from the numerical analysis agreed with the on-site observations.
For this study, we conducted a non-isothermal CFD analysis using data coming from the local meteorological bureau containing wind velocity, prevailing wind direction, and air temperature, and observed data containing ground and surface temperatures. The ground and building surface temperatures were calculated in step 2 based on the unsteady state heat balance calculation, including three-dimensional radiation and one-dimensional conduction calculations. Based on the results in these two steps, we once again conducted a CFD analysis to obtain more accurate results. Finally, we calculated SET*, a comprehensive index, to evaluate outdoor thermal comfort in the target area, which was in the centre of the analysis domain. Xi [31], Xuan et al. [32] and Zhou et al. [34] conducted their research using a similar prediction method.

Simulation Date
In this study, a weather database of China [35] was used. The typical meteorological day in summer is 1 July and the typical meteorological day in winter is 2 January (Table 1).

Simulation Time
We conducted an internet-based questionnaire survey in 2013 to determine the time of day when residents go outdoors in the residential district. More than 300 survey answers were collected. The question used in the survey is shown in Table 2.

Simulation Time
We conducted an internet-based questionnaire survey in 2013 to determine the time of day when residents go outdoors in the residential district. More than 300 survey answers were collected. The question used in the survey is shown in Table.2.  Figure 2 presents the survey results and weather data for 1 July of a typical year. The simulation time was set at 16:00 in the summer because of the high percentage of people who reported going outdoors (approximately 50%) and the high temperature (approximately 34 °C). Figure 3 presents the survey results for winter and the weather data on 2 January of a typical year. The simulation time was set to 9:00 in the winter because of the high percentage of people who go outdoors at that time (approximately 60%) and the low air temperature (approximately 4 °C).   We conducted an internet-based questionnaire survey in 2013 to determine the time of day when residents go outdoors in the residential district. More than 300 survey answers were collected. The question used in the survey is shown in Table.2.  Figure 2 presents the survey results and weather data for 1 July of a typical year. The simulation time was set at 16:00 in the summer because of the high percentage of people who reported going outdoors (approximately 50%) and the high temperature (approximately 34 °C). Figure 3 presents the survey results for winter and the weather data on 2 January of a typical year. The simulation time was set to 9:00 in the winter because of the high percentage of people who go outdoors at that time (approximately 60%) and the low air temperature (approximately 4 °C).   Figure 3 presents the survey results for winter and the weather data on 2 January of a typical year. The simulation time was set to 9:00 in the winter because of the high percentage of people who go outdoors at that time (approximately 60%) and the low air temperature (approximately 4 • C).

Analysis Method and Conditions
In this study, we used a numerical analysis system by integrating STAR-CD/RADX (CD adapco Group, Melville, NY USA) with additional codes. Figure 4 shows the flowchart of a coupled approach applied in this study. Our analysis was conducted in four steps. In step 1, we conducted a nonisothermal CFD analysis. We used meteorological conditions (air temperature, wind velocity, and prevailing wind direction) and ground and building surface temperatures observed in the environment as the initial and analysis conditions (Table 3).
In step 2, we conducted an unsteady state heat balance analysis for 16:00 on 1 July obtained from the ground and building surface temperatures. We included three-dimensional radiation and onedimensional conduction calculations in this process (Table 4). In step 3, we conducted a nonisothermal CFD analysis on account of the results of step 1 and step 2. The analysis conditions for step 3 are shown in Table 5. In step 4, we evaluated the thermal comfort for the study area with outcomes from step 3 (wind velocity, air temperature, humidity, and the mean radiant temperature (MRT)) and personal variables (activity and clothing). Table 6 lists the structures of the ground and buildings (wall and roof) and the corresponding thermal properties. Table 7 lists the surface properties of the ground, grass, building wall, and roof.

Analysis Method and Conditions
In this study, we used a numerical analysis system by integrating STAR-CD/RADX (CD adapco Group, Melville, NY USA) with additional codes. Figure 4 shows the flowchart of a coupled approach applied in this study. Our analysis was conducted in four steps. In step 1, we conducted a non-isothermal CFD analysis. We used meteorological conditions (air temperature, wind velocity, and prevailing wind direction) and ground and building surface temperatures observed in the environment as the initial and analysis conditions (Table 3).     In step 2, we conducted an unsteady state heat balance analysis for 16:00 on 1 July obtained from the ground and building surface temperatures. We included three-dimensional radiation and one-dimensional conduction calculations in this process (Table 4). In step 3, we conducted a non-isothermal CFD analysis on account of the results of step 1 and step 2. The analysis conditions for step 3 are shown in Table 5. In step 4, we evaluated the thermal comfort for the study area with outcomes from step 3 (wind velocity, air temperature, humidity, and the mean radiant temperature (MRT)) and personal variables (activity and clothing). Table 6 lists the structures of the ground and buildings (wall and roof) and the corresponding thermal properties. Table 7 lists the surface properties of the ground, grass, building wall, and roof.

Turbulence Model for Computational Fluid Dynamics (CFD) Analysis
A cubic non-linear k-ε model proposed by Craft et al. [36] is selected as the turbulence model for the CFD analysis. This turbulence model was originally developed to capture turbulence anisotropy by proposing a cubic relation between stress and strain, which significantly improves the predictions of three-dimensional fluid flow and heat transfer not only in and around a single object  A cubic non-linear k-ε model proposed by Craft et al. [36] is selected as the turbulence model for the CFD analysis. This turbulence model was originally developed to capture turbulence anisotropy by proposing a cubic relation between stress and strain, which significantly improves the predictions of three-dimensional fluid flow and heat transfer not only in and around a single object   [36] is selected as the turbulence model for the CFD analysis. This turbulence model was originally developed to capture turbulence anisotropy by proposing a cubic relation between stress and strain, which significantly improves the predictions of three-dimensional fluid flow and heat transfer not only in and around a single object  [36] is selected as the turbulence model for the CFD analysis. This turbulence model was originally developed to capture turbulence anisotropy by proposing a cubic relation between stress and strain, which significantly improves the predictions of three-dimensional fluid flow and heat transfer not only in and around a single object [36][37][38], but also around several objects [39]. This model has been successfully and widely used in mechanical engineering because of its high prediction accuracy. Its application in the field of wind climate around buildings is tested for the first time by our group. Specifically, its performance is compared with two other turbulence models (the standard k-ε model and the quadratic non-linear k-ε model proposed by Lien et al. [40]) and shows the best agreement with a wind tunnel experiment from the AIJ benchmark cases [41,42].
Li et al. [43] compared 15 kinds of turbulence models for outdoor wind environment numerical simulation using STAR-CD with the wind tunnel test provide by AIJ [41,44] and the results show that the Suga cubic non-linear k-ε model is of high precision for wind field solution.

Heat Transfer Analysis
To calculate urban surface temperatures, all the surfaces in the computational domain are divided into small surfaces. For each small surface, solar radiation, sky radiation, longwave radiation between it and other surfaces, convective heat transfer and latent heat transfer between it and the ambient air, and conduction heat transfer through it are considered. The shape factor is calculated using the Monte Carlo method [45,46] and the radiative heat transfer is calculated using Gebhart's absorption factor [47]. Given that the values of Gebhart's absorption factor are different because of the different absorptivities under shortwave and longwave radiations, they are calculated separately for each surface. According to previous studies [28,48,49], urban surface temperatures are well reproduced by considering the heat transfers mentioned above.

Analysis Model and Cases
The analysis model used in this study is shown in Figure 5. We determined the size of the computational domain based on the Architectural Institute of Japan Guidelines [50]. The size of computational domain in summer was 1102 m (x) * 1320 m (y) * 470 m (z), and the size in winter was 1342 m (x) * 1200 m (y) * 470 m (z). The grids of the computational domain in summer and in winter were 51(x) * 77(y) * 35(z) and 76(x) * 66(y) * 35(z), respectively. The sizes and grids of the residential area in summer and winter were same. We used six-floor residential buildings as the research object as they are the most common buildings in the study area. The density of the buildings was approximately 30% and the height of the first floor was 3.6 m; the height of other floors was 3 m.  [36][37][38], but also around several objects [39]. This model has been successfully and widely used in mechanical engineering because of its high prediction accuracy. Its application in the field of wind climate around buildings is tested for the first time by our group. Specifically, its performance is compared with two other turbulence models (the standard k-ε model and the quadratic non-linear k-ε model proposed by Lien et al. [40]) and shows the best agreement with a wind tunnel experiment from the AIJ benchmark cases [41,42]. Li et al. [43] compared 15 kinds of turbulence models for outdoor wind environment numerical simulation using STAR-CD with the wind tunnel test provide by AIJ [41,44] and the results show that the Suga cubic non-linear k-ε model is of high precision for wind field solution.

Heat Transfer Analysis
To calculate urban surface temperatures, all the surfaces in the computational domain are divided into small surfaces. For each small surface, solar radiation, sky radiation, longwave radiation between it and other surfaces, convective heat transfer and latent heat transfer between it and the ambient air, and conduction heat transfer through it are considered. The shape factor is calculated using the Monte Carlo method [45,46] and the radiative heat transfer is calculated using Gebhart's absorption factor [47]. Given that the values of Gebhart's absorption factor are different because of the different absorptivities under shortwave and longwave radiations, they are calculated separately for each surface. According to previous studies [28,48,49], urban surface temperatures are well reproduced by considering the heat transfers mentioned above.

Analysis Model and Cases
The analysis model used in this study is shown in Figure 5. We determined the size of the computational domain based on the Architectural Institute of Japan Guidelines [50]. The size of computational domain in summer was 1102 m (x) * 1320 m (y) * 470 m (z), and the size in winter was 1342 m (x) * 1200 m (y) * 470 m (z). The grids of the computational domain in summer and in winter were 51(x) * 77(y) * 35(z) and 76(x) * 66(y) * 35(z), respectively. The sizes and grids of the residential area in summer and winter were same. We used six-floor residential buildings as the research object as they are the most common buildings in the study area. The density of the buildings was approximately 30% and the height of the first floor was 3.6 m; the height of other floors was 3 m.  As seen in Figure 6, four different piloti arrangements were considered in the study area to determine the optimal arrangement (Table 8). Then, we investigated the relationship between the piloti ratio and the wind environment based on the optimal piloti arrangement. As seen in Figure 6, four different piloti arrangements were considered in the study area to determine the optimal arrangement (Table 8). Then, we investigated the relationship between the piloti ratio and the wind environment based on the optimal piloti arrangement. Figure 6. The piloti arrangements analysed for the study area.

Surface Temperature
The surface temperatures of Case 0-S are presented in Figure 7a. For the building, the temperature of the roof surface was the highest, reaching 46 °C, followed by the west wall (approximately 37 °C). The temperature of the north wall was the lowest (approximately 32 °C). The direction of sunshine in the summer case is also shown in the figure.
The surface temperatures of Case 0-W are presented in Figure 7b. In this case, the sunshine reached the 6th or 5th floor. For the building, the east wall had the highest temperature (approximately 18 °C). Most of the ground was in shadow. The temperature of the ground was approximately 32 °C. The wind environment was more important than the thermal environment in the winter case because most of the building and ground were in shadow. Therefore, to save time, the MRTs of the winter cases were not calculated.     When the piloti ratio was 0 (Case 0-S) or the piloti was in the middle of the building (Case 2-S), the portion of wind velocity less than 0.5 m/s occupied about half of the cumulative distribution. When the piloti was at the two ends of the building (Case 4-S), the wind velocities greater than 0.5 m/s were over 75% of the cumulative distribution. Case 4-S significantly improved the wind environment. Therefore, we concluded that the piloti at the two ends of the building was the optimal piloti arrangement. The left part of Figure 8b shows the distributions of wind velocities at 1.5 m for Case 0-W to Case 4-W. The wind path formed along the east to west direction. After adding the piloti space, the wind path weakened. The right part of Figure 8b presents the probability density and cumulative distribution of the wind velocities at 1.5 m of non-piloti space. After adding piloti, the maximum wind velocity decreased. From Case 1-W to Case 4-W, all the average wind velocities were lower than Case 0.   The surface temperatures of Case 0-W are presented in Figure 7b. In this case, the sunshine reached the 6th or 5th floor. For the building, the east wall had the highest temperature (approximately 18 • C). Most of the ground was in shadow. The temperature of the ground was approximately 32 • C. The wind environment was more important than the thermal environment in the winter case because most of the building and ground were in shadow. Therefore, to save time, the MRTs of the winter cases were not calculated. Figure 8a presents the distributions of wind velocity at 1.5 m for Case 0-S to Case 4-S, and it shows the probability density and cumulative distribution of wind velocity for Case 0-S to Case 4-S. When the piloti ratio was 0 (Case 0-S) or the piloti was in the middle of the building (Case 2-S), the portion of wind velocity less than 0.5 m/s occupied about half of the cumulative distribution. When the piloti was at the two ends of the building (Case 4-S), the wind velocities greater than 0.5 m/s were over 75% of the cumulative distribution. Case 4-S significantly improved the wind environment. Therefore, we concluded that the piloti at the two ends of the building was the optimal piloti arrangement. The left part of Figure 8b shows the distributions of wind velocities at 1.5 m for Case 0-W to Case 4-W. The wind path formed along the east to west direction. After adding the piloti space, the wind path weakened. The right part of Figure 8b presents the probability density and cumulative distribution of the wind velocities at 1.5 m of non-piloti space. After adding piloti, the maximum wind velocity decreased. From Case 1-W to Case 4-W, all the average wind velocities were lower than Case 0.  When the piloti ratio was 0 (Case 0-S) or the piloti was in the middle of the building (Case 2-S), the portion of wind velocity less than 0.5 m/s occupied about half of the cumulative distribution. When the piloti was at the two ends of the building (Case 4-S), the wind velocities greater than 0.5 m/s were over 75% of the cumulative distribution. Case 4-S significantly improved the wind environment. Therefore, we concluded that the piloti at the two ends of the building was the optimal piloti arrangement. The left part of Figure 8b shows the distributions of wind velocities at 1.5 m for Case 0-W to Case 4-W. The wind path formed along the east to west direction. After adding the piloti space, the wind path weakened. The right part of Figure 8b presents the probability density and cumulative distribution of the wind velocities at 1.5 m of non-piloti space. After adding piloti, the maximum wind velocity decreased. From Case 1-W to Case 4-W, all the average wind velocities were lower than Case 0.    This study gave priority to the improvement of a thermal environment in summer because (i) Wuhan is famous for its hot and humid summer in China, and the climate problem is more serious in summer and because (ii) in cold-winter days people can easily improve their outdoor thermal comfort by clothing, warm paste and exercise, while in hot-summer days, the effect of such methods is very limited. In summer, the piloti arrangement of case 4 shows the best effect in wind environment ( Figure 8a) and SET* (Figure 9b), and in winter, the arrangement of case 4 can decrease the average wind velocity although it was not the best (Figure 8b). However, the optimal piloti arrangement is determined by the thermal environment in summer as the objective of the design is to solve the heat stress problem. Although there could be some improvements on the MRT and SET* in winter by altering the piloti arrangement, it will have no impact on the conclusion in this paper.

Wind Velocity
The upper part of Figure 10a shows the distributions of wind velocities at 1.5 m for Case 0-S to Case 4-80-S. The wind path formed along a south to north direction. With an increasing piloti ratio, the wind path increased. The wind velocities beside the south and north side of the building were very weak, and the enlarged wind path improved the wind velocity of this area. The lower part of Figure 10a lists the average wind velocities of different summer cases. The wind environment was improved with the increasing piloti ratio in the summer. The wind velocity in the red dashed areas (weak wind areas) changed significantly.
The upper part of Figure 10b shows the distributions of wind velocities at 1.5 m for Case 0-W to Case 4-80-W. The wind path formed in the east to west direction. With the increasing piloti ratio, the wind path weakened. The lower part of Figure 10b lists the average wind velocities of different cases in winter. The wind environment changed slightly in winter with the increasing piloti ratio. With the increasing piloti ratio, the average wind velocity first decreased and then increased; with the increasing piloti ratio, the wind path weakened and the wind velocity decreased. On the first floor, the building density decreased and the wind velocity increased. This study gave priority to the improvement of a thermal environment in summer because (i) Wuhan is famous for its hot and humid summer in China, and the climate problem is more serious in summer and because (ii) in cold-winter days people can easily improve their outdoor thermal comfort by clothing, warm paste and exercise, while in hot-summer days, the effect of such methods is very limited. In summer, the piloti arrangement of case 4 shows the best effect in wind environment ( Figure 8a) and SET* (Figure 9b), and in winter, the arrangement of case 4 can decrease the average wind velocity although it was not the best (Figure 8b). However, the optimal piloti arrangement is determined by the thermal environment in summer as the objective of the design is to solve the heat stress problem. Although there could be some improvements on the MRT and SET* in winter by altering the piloti arrangement, it will have no impact on the conclusion in this paper.

Wind Velocity
The upper part of Figure 10a shows the distributions of wind velocities at 1.5 m for Case 0-S to Case 4-80-S. The wind path formed along a south to north direction. With an increasing piloti ratio, the wind path increased. The wind velocities beside the south and north side of the building were very weak, and the enlarged wind path improved the wind velocity of this area. The lower part of Figure 10a lists the average wind velocities of different summer cases. The wind environment was improved with the increasing piloti ratio in the summer. The wind velocity in the red dashed areas (weak wind areas) changed significantly.  The average wind velocity of one month was much higher than the wind velocity for the analysis time. To avoid wind-induced discomfort, it is better to consider the average wind velocity for one month. Based on the input data, we calculated the wind velocity at a height of 1.5 m (0.78 m/s in the summer and 1.24 m/s in the winter). According to the definition of the wind velocity ratio, the relationship between the piloti ratio and the average wind velocity ratio can be represented as According to a weather database of China [35], the average wind velocity in July (the hottest month) is 3.6 m/s at the height of 10 m, and the average wind velocity in January (the coldest month) is 2.4 m/s at the height of 10 m. According to the input data (ground roughness α = 0.25), we calculated that the average wind velocities at the height of 1.5 m in summer and winter are 2.24 m/s and 1.49 m/s, respectively. The main part of Figure 11 was generated from the data in Table 9. Table 10 presents the relationship between the piloti ratio and the average wind velocity of the non-piloti area and the red dashed area in summer and in winter.  The upper part of Figure 10b shows the distributions of wind velocities at 1.5 m for Case 0-W to Case 4-80-W. The wind path formed in the east to west direction. With the increasing piloti ratio, the wind path weakened. The lower part of Figure 10b lists the average wind velocities of different cases in winter. The wind environment changed slightly in winter with the increasing piloti ratio. With the increasing piloti ratio, the average wind velocity first decreased and then increased; with the increasing piloti ratio, the wind path weakened and the wind velocity decreased. On the first floor, the building density decreased and the wind velocity increased.
The average wind velocity of one month was much higher than the wind velocity for the analysis time. To avoid wind-induced discomfort, it is better to consider the average wind velocity for one month. Based on the input data, we calculated the wind velocity at a height of 1.5 m (0.78 m/s in the summer and 1.24 m/s in the winter). According to the definition of the wind velocity ratio, the relationship between the piloti ratio and the average wind velocity ratio can be represented as R i is the wind velocity ratio. u i is the wind speed at point i (m/s) when buildings exist. u i0 is the wind speed at point i (m/s) when buildings do not exist, usually the same as the inflow value at the same height.
According to a weather database of China [35], the average wind velocity in July (the hottest month) is 3.6 m/s at the height of 10 m, and the average wind velocity in January (the coldest month) is 2.4 m/s at the height of 10 m. According to the input data (ground roughness α = 0.25), we calculated that the average wind velocities at the height of 1.5 m in summer and winter are 2.24 m/s and 1.49 m/s, respectively. The main part of Figure 11 was generated from the data in Table 9. Table 10 presents the relationship between the piloti ratio and the average wind velocity of the non-piloti area and the red dashed area in summer and in winter.   Figure 11. The criteria for assessing the wind-induced discomfort considering the temperature effect.
The main content of Figure 11 was developed from Murakami's research [51]. According to a weather database of China [35], the average temperature in July (the hottest month) is 29.6 °C and the average temperature in January (the coldest month) is 3.8 °C. As seen in Figure 11, if the wind velocity is stronger than 2.8 m/s or weaker than 1.24 m/s, it will cause wind-induced discomfort in summer; if the wind velocity is stronger than 1.9 m/s, it will cause wind-induced discomfort in winter. Figure 12 shows that the piloti ratio should be between 12% and 38% to avoid wind-induced discomfort. Figure 12. The relationship between the piloti ratio and the average wind velocity of non-piloti areas.

MRT
The distributions of MRT for Case 0-S to Case 4-80-S are displayed in the upper part of Figure  13a. With the increasing piloti ratio, the MRT decreased. A shaded space formed under the piloti. In Case 0-S, the temperature of the west wall on the first floor was high, so the MRT beside the west wall was very high. After adding piloti, the west wall on the first floor was under the piloti, so the The main content of Figure 11 was developed from Murakami's research [51]. According to a weather database of China [35], the average temperature in July (the hottest month) is 29.6 • C and the average temperature in January (the coldest month) is 3.8 • C. As seen in Figure 11, if the wind velocity is stronger than 2.8 m/s or weaker than 1.24 m/s, it will cause wind-induced discomfort in summer; if the wind velocity is stronger than 1.9 m/s, it will cause wind-induced discomfort in winter. Figure 12 shows that the piloti ratio should be between 12% and 38% to avoid wind-induced discomfort.  Figure 11. The criteria for assessing the wind-induced discomfort considering the temperature effect.
The main content of Figure 11 was developed from Murakami's research [51]. According to a weather database of China [35], the average temperature in July (the hottest month) is 29.6 °C and the average temperature in January (the coldest month) is 3.8 °C. As seen in Figure 11, if the wind velocity is stronger than 2.8 m/s or weaker than 1.24 m/s, it will cause wind-induced discomfort in summer; if the wind velocity is stronger than 1.9 m/s, it will cause wind-induced discomfort in winter. Figure 12 shows that the piloti ratio should be between 12% and 38% to avoid wind-induced discomfort. Figure 12. The relationship between the piloti ratio and the average wind velocity of non-piloti areas.

MRT
The distributions of MRT for Case 0-S to Case 4-80-S are displayed in the upper part of Figure  13a. With the increasing piloti ratio, the MRT decreased. A shaded space formed under the piloti. In Case 0-S, the temperature of the west wall on the first floor was high, so the MRT beside the west wall was very high. After adding piloti, the west wall on the first floor was under the piloti, so the temperature of the west wall in the first floor was less, and the MRT beside the west wall was low

MRT
The distributions of MRT for Case 0-S to Case 4-80-S are displayed in the upper part of Figure 13a. With the increasing piloti ratio, the MRT decreased. A shaded space formed under the piloti. In Case 0-S, the temperature of the west wall on the first floor was high, so the MRT beside the west wall was very high. After adding piloti, the west wall on the first floor was under the piloti, so the temperature of the west wall in the first floor was less, and the MRT beside the west wall was low because of the shaded space under the piloti. The lower part of Figure 13a shows the probability density and cumulative distribution of the MRT of the non-piloti area.
shows the probability density and cumulative distribution of MRT (non-piloti area) in the winter. Although the average MRT changed slightly with the increasing piloti ratio, the average MRT first increased, then decreased, and then increased again. The area that received direct sunlight increased with the increasing piloti ratio and this caused the MRT to increase. It was difficult to receive diffuse radiation from the sky under the west piloti, so its MRT was lower than the surrounding space, and this decreased the surrounding MRT.

SET*
The upper part of Figure 14a shows the distributions of SET* in the summer cases. Piloti can provide shaded places and this can rapidly decrease the SET*. The SET* was very high beside the south and north sides of the building because the wind velocity was very weak. The SET* was lower beside the east and west sides of the building because of the wind path. The lower part of Figure 14a presents the probability density and cumulative distribution of SET* (non-piloti area) for the summer cases. The minimum SET* value was approximately 31 °C, and the maximum SET* value was approximately 42 °C.  The upper part of Figure 14b presents the distributions of SET* for Case 0-W to Case 4-80-W. The SET* value of the area that received direct sunshine was highest and most of the area was in shadow. The SET* beside the two ends of the building was higher than that of other areas because the wind velocity besides the two ends of the building was very weak. The SET* changed slightly after adding piloti. The lower part of Figure 14b displays the probability density and cumulative distribution of SET* of the non-piloti area for the winter cases. The minimum SET* value was approximately 16 °C, and the maximum SET* value was approximately 29 °C. With the increasing piloti ratio, the average SET* first increased, then decreased, and then increased again. This trend was  Figure 13b. In the winter, the MRT changed slightly after adding piloti. The lower part of Figure 13b shows the probability density and cumulative distribution of MRT (non-piloti area) in the winter. Although the average MRT changed slightly with the increasing piloti ratio, the average MRT first increased, then decreased, and then increased again. The area that received direct sunlight increased with the increasing piloti ratio and this caused the MRT to increase. It was difficult to receive diffuse radiation from the sky under the west piloti, so its MRT was lower than the surrounding space, and this decreased the surrounding MRT.

SET*
The upper part of Figure 14a shows the distributions of SET* in the summer cases. Piloti can provide shaded places and this can rapidly decrease the SET*. The SET* was very high beside the south and north sides of the building because the wind velocity was very weak. The SET* was lower beside the east and west sides of the building because of the wind path. The lower part of Figure 14a presents the probability density and cumulative distribution of SET* (non-piloti area) for the summer cases. The minimum SET* value was approximately 31 • C, and the maximum SET* value was approximately 42 • C. The probability density and cumulative distribution of SET* (non-piloti area).
The probability density and cumulative distribution of SET* (non-piloti area).

Discussion
In this paper, case studies of a residential area with different piloti arrangements and the ratio by numerical simulation were carried out, and the piloti showed good passive climate adaptability in a hot summer and cold winter city. The relationships between piloti variables and outdoor thermal environment (T, V, MRT, SET*) of the residential area both in winter and in summer in Wuhan is studied for the first time. Data in Table 6 can be converted to the relationship between piloti ratio and wind speed in other cities with similar climate with Wuhan, which shows a certain adaptability.
Although there are some interesting results revealed by this paper, there are also limitations. First, we only studied the most basic and common determinant that arranged the residential area, while other arrangement forms (point, enclosed) with different building shapes also accounted for a large proportion. Second, we only consider the piloti variable, however, the improvement of the thermal environment is not limited by a single piloti variable. Other passive climate adaptability elements, such as reflective materials and the green wall should also be taken into account. Third, Trees, pedestrians, and cars are omitted in the simulation. Fourth, only a typical meteorological day is considered in this study, other typical days such as the hottest day and the coldest day could also be taken into account.
In our future work, the thermal environment of residential areas under different arrangement forms (point, enclosed) will be researched. The upper part of Figure 14b presents the distributions of SET* for Case 0-W to Case 4-80-W. The SET* value of the area that received direct sunshine was highest and most of the area was in shadow. The SET* beside the two ends of the building was higher than that of other areas because the wind velocity besides the two ends of the building was very weak. The SET* changed slightly after adding piloti. The lower part of Figure 14b displays the probability density and cumulative distribution of SET* of the non-piloti area for the winter cases. The minimum SET* value was approximately 16 • C, and the maximum SET* value was approximately 29 • C. With the increasing piloti ratio, the average SET* first increased, then decreased, and then increased again. This trend was likely due to the changing trends of wind velocity and the MRT. The SET* values changed slightly in winter with the increasing piloti ratio.

Discussion
In this paper, case studies of a residential area with different piloti arrangements and the ratio by numerical simulation were carried out, and the piloti showed good passive climate adaptability in a hot summer and cold winter city. The relationships between piloti variables and outdoor thermal environment (T, V, MRT, SET*) of the residential area both in winter and in summer in Wuhan is studied for the first time. Data in Table 6 can be converted to the relationship between piloti ratio and wind speed in other cities with similar climate with Wuhan, which shows a certain adaptability.
Although there are some interesting results revealed by this paper, there are also limitations. First, we only studied the most basic and common determinant that arranged the residential area, while other arrangement forms (point, enclosed) with different building shapes also accounted for a large proportion. Second, we only consider the piloti variable, however, the improvement of the thermal environment is not limited by a single piloti variable. Other passive climate adaptability elements, such as reflective materials and the green wall should also be taken into account. Third, Trees, pedestrians, and cars are omitted in the simulation. Fourth, only a typical meteorological day is considered in this study, other typical days such as the hottest day and the coldest day could also be taken into account.
In our future work, the thermal environment of residential areas under different arrangement forms (point, enclosed) will be researched.

Conclusions
We studied the outdoor thermal environment in the hottest month (July) and coldest month (January) in Wuhan. We found that setting piloti at the two ends of the building was the best piloti arrangement for climate adaptation. It improved the wind environment and thermal environment.
Based on the optimal piloti arrangement, we studied the relationship between the piloti ratio (0%, 20%, 40%, 60%, and 80%) and the outdoor wind environment. In the summer, the wind velocity increased with the increasing piloti ratio, as the piloti could strengthen the wind path (from south to north). In the winter the wind environment changed slightly with the increasing piloti ratio because the piloti had little effect on the wind path (from east to west). The criteria for assessing wind-induced discomfort considering the temperature effects were used to evaluate the wind environment. We found that the piloti ratio should be between 12% and 38% to avoid wind-induced discomfort.
In the summer, with the increasing piloti ratio, MRT and SET* decreased. The wind path significantly affected the SET* value. MRT and SET* changed slightly with the increasing piloti ratio because most of the building and ground were in shadow in the winter cases.