Impact of Indoor-Outdoor Temperature Difference on Building Ventilation and Pollutant Dispersion within Urban Communities

: Mechanical ventilation consumes a huge amount of global energy. Natural ventilation is a crucial solution for reducing energy consumption and enhancing the capacity of atmospheric self-puriﬁcation. This paper evaluates the impacts of indoor-outdoor temperature differences on building ventilation and indoor-outdoor air pollutant dispersion in urban areas. The Computational Fluid Dynamics (CFD) method is employed to simulate the ﬂow ﬁelds in the street canyon and indoor environment. Ventilation conditions of single-side ventilation mode and cross-ventilation mode are investigated. Air change rate, normalized concentration of trafﬁc-related air pollutant (CO), intake fraction and exposure concentration are calculated to for ventilation efﬁciency investigation and exposure assessment. The results show that cross ventilation increases the air change rate for residential buildings under isothermal conditions. With the indoor-outdoor temperature difference, heating could increase the air change rate of the single-side ventilation mode but restrain the capability of the cross-ventilation mode in part of the ﬂoors. Heavier polluted areas appear in the upstream areas of single-side ventilation modes, and the pollutant can diffuse to middle-upper ﬂoors in cross-ventilation modes. Cross ventilation mitigates the environmental health stress for the indoor environment when indoor-outdoor temperature difference exits and the personal intake fraction is decreased by about 66% compared to the single-side ventilation. Moreover, the existence of indoor-outdoor temperature differences can clearly decrease the risk of indoor personal exposure under both two natural ventilation modes. The study numerically investigates the building ventilation and pollutant dispersion in the urban community with natural ventilation. The method and the results are helpful references for optimizing the building ventilation plan and improving indoor air quality.


Introduction
More than half of the current global population lives in urban areas. This proportion keeps increasing and could reach 68% by 2050 [1]. Rapid urbanization, deteriorating air quality, and air pollutants exposure increase the risks of respiratory and cardiovascular diseases on residents in urban areas [2,3]. However, pollutants dispersion and human exposure in urban areas are closely linked to urban ventilation and building ventilation [4,5], especially for the buildings that are naturally ventilated [6][7][8]. Indoor activities occupy about 90% of the majority of people's lifetimes, and indoor air quality is affected by outdoor air pollutants via ventilation. Hence, the residents living in curb-side buildings are exposed of building ventilation modes are considered for different scenarios. Air change rate, normalized concentration, and exposure concentration in different scenarios are analysed for numerical investigation of the pollutant dispersion and the related health effects. This work applies a useful method and helpful references for optimizing the program of building ventilation and improving indoor air quality. In this paper, the Ansys FLUENT (version 15.0) with RNG k-ε model is applied for airflow simulations. The flow in the modelling is assumed to be an incompressible fluid along with the Boussinesq approximation [64].

Materials and Methods
Due to the inertness and the low background concentration in the atmosphere, CO is widely used as a tracer for the traffic emission in simulating the dispersion of near-ground vehicular pollutant and investigating related human exposure [5,61,63]. In this work, we also use CO as the indication for traffic-related air pollutants. The governing equation for CO dispersion is: where C is the CO concentration (kg/m 3 ). D m and D t are the molecular diffusivity and the turbulent diffusivity. D t = ν t /S C t , where ν t and S C t are the kinematic eddy viscosity and the turbulent Schmidt number. S C t = 0.7 is used in this work as we find it has the best performance in our preliminary work [65] comparing to the wind-tunnel experiment results [52]. S is the emission rate of CO and is set as 1.3 × 10 −6 kg/m 3 /s referring to [12].

Boundary Conditions
The principles of AIJ (Architectural Institute of Japan) and COST (European Cooperation in Science and Technology) are followed in this study for the model set-up of the boundary condition [66,67]. Non-slip wall conditions and standard wall function are employed for near-wall and near-surface treatments (Figure 1). Zero normal gradient conditions are applied at two domain laterals (i.e., symmetry), domain top (i.e., symmetry), and the domain outlet (i.e., outflow). The outdoor wind velocity above the ground surface increases with the rising height due to the existence of the surface friction. Hence the wind velocity under 300 m is defined by the mean gradient of the velocity. The profile of the wind velocity is described by Equation (2) [4,68]. The outdoor wind velocity above the ground surface increases with the rising height due to the existence of the surface friction. Hence the wind velocity under 300 m is defined by the mean gradient of the velocity. The profile of the wind velocity is described by Equation (2) [4,68].
where U z is the wind velocity at the height of z; U g is the wind velocity above the atmospheric boundary layer; h is the height of the ABL, and we use h = 270 m in this work; α is the underlying surface roughness, and in this work α = 0.22 is adopted to denote the moderate-dense urban area. The outdoor incoming flow at the domain inlet is set as the inlet velocity, and the temperature of the inlet flow is set as 295.15 K. The SIMPLE scheme is applied for coupling pressure and velocity.

Model Set-Up and Mesh Arrangement
A single-side ventilation model and cross ventilation model are set up for studying their impacts on natural building ventilation as well as their influences on the diffusion of the outdoor air pollutants to indoor environments. Figure 2a illustrations the full-scale street canyon models employed in this study. The 2D street canyons consist of five uniform residential building models and four canyons, with the building height (H) = 24 m and the street width (W) = 24 m. The aspect ratio (AR = H/W) is 1. The target area is the third street canyon and the buildings on the two sides. The CO emission source is set in the target street canyon with 0.5 m in height and 16 m in width. The distances of the emission source to the buildings are set as 4 m on both sides. The two target buildings are named as Building 1 and Building 2, as shown in Figure 2a. Other buildings at the upstream and the downstream are employed for representing the impacts of the roughness elements and getting well-developed turbulence [54][55][56]69].
The dimension of the building model is 20 m (x) × 12 m (y) × 24 m (z). Each building has eight floors, and each floor has two rooms with the size of 6 m (x)× 4 m (y)× 2.7 m (z). Each room on the first floor (F1) has two doors (1.6 m × 2 m) on the facade and the opposite side. Each room on the second floor and above floors (F2-F8) has one window (4 m × 1 m) on the facade facing towards the street, and a door (1.6 m × 2 m) on the opposite side of the window. The space in the middle of each floor is the public corridor. In single-side ventilation scenarios, the doors or windows facing the street are open and the doors on the opposite side are closed. In cross-ventilation scenarios, all windows and doors are open. Simulation case series are named as listed in Table 1 according to the ventilation modes and indoor-outdoor air temperature differences (∆T).   Table 1 according to the ventilatio modes and indoor-outdoor air temperature differences (ΔT). Non-slip wall conditions and standard wall functions are applied for all groun surfaces, building walls and indoor walls. In cases with ΔT ≠ 0, the indoor thermal sour is supposed to be homogeneous on the six room walls for each room, and the heat flux set as 100 W/m 2 . The dimension of the computational domain is 196 m × 12 m × 150 (>>5H). Structured hexahedral mesh is applied in the whole computational domain, an dense mesh is adopted at each vent (Figure 2b,c). The total grid numbers produced f single-side ventilation and cross ventilation modes are 970,000 and 1,540,000, respectivel Non-slip wall conditions and standard wall functions are applied for all ground surfaces, building walls and indoor walls. In cases with ∆T = 0, the indoor thermal source is supposed to be homogeneous on the six room walls for each room, and the heat flux is set as 100 W/m 2 . The dimension of the computational domain is 196 m × 12 m × 150 m (>>5H). Structured hexahedral mesh is applied in the whole computational domain, and dense mesh is adopted at each vent (Figure 2b,c). The total grid numbers produced for single-side ventilation and cross ventilation modes are 970,000 and 1,540,000, respectively.

Comparison with the Wind-Tunnel Experiments for the Flow and Dispersion
As reviewed by Toparlar et al. [39], most CFD studies did not include a direct validation study since the high-quality observation data were very limited on both quantity and quality. Moreover, the uncontrolled or very limited controlled boundary conditions were big challenges for the repetition tests. Hence wind-tunnel experiments were widely employed for the comparison in CFD studies to verify the model reliability when the Reynolds number (Re) independence was satisfied (Re >> 11,000) [70][71][72]. In our preliminary work [73], we implemented a series of detailed comparisons for the RNG k-ε model employed in this work. The dispersion model applied was also validated in our earlier work [72]. Some important results are presented in Figures A1 and A2 in Appendix A. The parameter air change per hour (ACH) is applied to evaluate the effect of the natural ventilation in the room [29]. Equations (3) and (4) are the definition of ACH.
Q is the airflow rate of the room in cubic meters per second. v is the airflow velocity perpendicular to the vent (i.e., windows and doors in this work). V denotes the bulk volume of the room, and S is the cross-section area of the vent (i.e., windows and doors in this work).

Normalized Concentration (K) and Exposure Concentration (E)
In the real world, the emission strengths of different sources are not the same and may even be at different orders in various kinds of street networks. For the purpose of setting up a universal methodology on pollutants and exposure assessment, the normalized concentration K is employed to better illustrate the deviation for the target research area. The influence caused by different orders of the emission strengths could be eliminated by using K [74]. K and the exposure concentration E are defined by Equations (5) and (6).
K is the normalized CO concentration (kg/m). C is the CO concentration (kg/m 3 ). U ref is the reference velocity (m/s). H is the building height (m). L is the length of the emission source (m). S is the emission rate of the source (m/s). E is the exposure concentration (kg·s/m 3 ). ∆t is the exposure time (s). i, j, and k are age group, environment category and building room, respectively.

Intake Fraction (IF) and Personal Intake Fraction (P_IF)
Intake fraction (IF) is an important index to assess the inhalation exposure of air pollutants for a certain population [5,61]. Hang et al. [5] optimized it as personal intake fraction (P_IF) for setting up a universal evaluation method. This normalized exposure index P_IF is independent of the population size, density, and the air pollution level. It is widely used to quantitatively evaluate the exposure of traffic-related air pollutants for indoor environments or outdoor environments at street-scale (~100 m) [62,63,75].
Equations (7) and (8) define IF and P_IF, respectively. n is the number of age groups of the assessed population. M is the number of investigated micro-environment categories. P i is the number of persons in the certain age group i. Br i,j (m 3 /s) denotes the breathing rate in volume mean for individuals of group i in environment j. ∆t is the exposure time of group i in environment j. Ce j (kg/m 3 ) is the temporal mean concentration of the certain pollutant in environment j. m (kg) is the total of the certain pollutant throughout the duration.   When ∆T = 0, the flow field in the target canyon is illustrated in Figure 3b. The expanded main vortex occupies the whole canyon. The air mass spills over the canyon edge and is transported downstream at the upper layer. Thermal-driven upstream enhances the updraft on the leeward side, and the largest velocity could reach 1.6 m/s. Accordingly, the downdraft on the windward side is enhanced, as well. Despite the sinking being partly offset by the thermal buoyancy, the sinking velocity is still lager than 0.8 m/s. Consequently, a downward cascading is generated to the windward building and an upward cascading is produced to the leeward building. Cascading effect describes the phenomenon that air mass flowing in between the neighbouring floors. Figure 4 displays the detailed flow of an upward cascading effect in the upstream building (canyon 2) as an example. Figure 3c,d depict the streamline and flow in the upstream canyon (canyon 2) and the downstream canyon (canyon 4). A significant upward cascading is observed in Figure 3c. Since the indoor heating of the right-side building, the indoor air mass slops over the room and flows into the room upstairs via the front windows. The velocity of this updraft is larger than 0.8 m/s. Functioning together with the downdraft of the main vortex is a sub vortex. The flow field in canyon 4 ( Figure 3d) is similar as that in the target canyon. However, the absence of the indoor heating of the right-side building leads to a large velocity of the downdraft, which could reach 1.4 m/s. Comparing to the downdraft on the windward in Figure 3b, the thermal-driven upstream cases by the indoor-outdoor air temperature difference in case [single-sided, 100] reaches 0.6 m/s.
Equations (7) and (8) define IF and P_IF, respectively. n is the number of age groups of the assessed population. M is the number of investigated micro-environment categories. Pi is the number of persons in the certain age group i. Bri,j (m 3 /s) denotes the breathing rate in volume mean for individuals of group i in environment j. ∆t is the exposure time of group i in environment j. Cej (kg/m 3 ) is the temporal mean concentration of the certain pollutant in environment j. m (kg) is the total of the certain pollutant throughout the duration.    When ΔT ≠ 0, the flow field in the target canyon is illustrated in Figure 3b. The expanded main vortex occupies the whole canyon. The air mass spills over the canyon edge and is transported downstream at the upper layer. Thermal-driven upstream enhances the updraft on the leeward side, and the largest velocity could reach 1.6 m/s. Accordingly, the downdraft on the windward side is enhanced, as well. Despite the sinking being partly offset by the thermal buoyancy, the sinking velocity is still lager than 0.8 m/s. Consequently, a downward cascading is generated to the windward building and an upward cascading is produced to the leeward building. Cascading effect describes the phenomenon that air mass flowing in between the neighbouring floors. Figure 4 displays the detailed flow of an upward cascading effect in the upstream building (canyon 2) as an example. Figure 3c,d depict the streamline and flow in the upstream canyon (canyon 2) and the downstream canyon (canyon 4). A significant upward cascading is observed in Figure 3c. Since the indoor heating of the right-side building, the indoor air mass slops over the room and flows into the room upstairs via the front windows. The velocity of this updraft is larger than 0.8 m/s. Functioning together with the downdraft of the main vortex is a sub vortex. The flow field in canyon 4 ( Figure 3d) is similar as that in the target canyon. However, the absence of the indoor heating of the right-side building leads to a large velocity of the downdraft, which could reach 1.4 m/s. Comparing to the downdraft on the windward in Figure 3b, the thermal-driven upstream cases by the indoor-outdoor air temperature difference in case [single-sided, 100] reaches 0.6 m/s.

Cross Ventilation Scenario
The streamline and flow in street canyons with cross-ventilated buildings are illustrated in Figure 5. With idealized isothermal conditions (Figure 5a), the flow pattern in the target canyon is a single-vortex type. The vortex occupies the full canyon, and its upper edge is at the same height with the canyon top. The air mass can flow across neighbouring canyons through the top floors of buildings owing to the cross ventilation. The velocity of the indoor airflow in the top floors of the buildings is significantly increased. Nevertheless, the influence on the airflow in the lower floors is negligible. Figure 5b presents the streamline and flow in the target canyon. As the results of the crossventilation mode and the downdraft in canyon 2 displayed in Figure 5b and c, the air mass

Cross Ventilation Scenario
The streamline and flow in street canyons with cross-ventilated buildings are illustrated in Figure 5. With idealized isothermal conditions (Figure 5a), the flow pattern in the target canyon is a single-vortex type. The vortex occupies the full canyon, and its upper edge is at the same height with the canyon top. The air mass can flow across neighbouring canyons through the top floors of buildings owing to the cross ventilation. The velocity of the indoor airflow in the top floors of the buildings is significantly increased. Nevertheless, the influence on the airflow in the lower floors is negligible. Figure 5b presents the streamline and flow in the target canyon. As the results of the cross-ventilation mode and the downdraft in canyon 2 displayed in Figure 5b,c, the air mass in canyon 2 can flow into the target canyon through each building floor. The spilling indoor air mass from each floor slops upwards on both buildings in the target canyon, converging to a main updraft and flowing to the lower reaches on the upper layer of the canyon. Owing to the large upward velocity on both sides, no downdraft forms in the target canyon. Therefore, the vortex in the target canyon fails to construct. Figure 5c,d presents the streamline and flow in the canyons' upstream (canyon 2) and downstream (canyon 4), respectively. The flow fields in these two canyons are both single-vortex patterns. The downdraft in canyon 2 flows into the indoor environment via the front windows, crosses the building horizontally via doors and corridors, and flows out via the opposite windows of the building. Corridors and the low layer in rooms are the areas with large velocity. Static wind areas are formed in each room above the height of the window. Though the flow pattern in canyon 4 is similar as that in canyon 2, the lack of heating on the windward wall increases the velocity of the downdraft. The biggest sinking velocity is larger than 1.8 m/s, and the velocity at the vortex edge is about 1.1 m/s.

Investigation for the Natural Ventilation Efficiency of the Canyon Buildings
The factor ACH as defined in Section 2.

Investigation for the Natural Ventilation Efficiency of the Canyon Buildings
The factor ACH as defined in Section 2.2.1 is used for investigating the efficiency of the target buildings with two ventilation modes. In the single-side ventilation mode, the proportion of the open windows is supposed to be at 100%, and all the doors in buildings are closed. In the assumption of the cross-ventilation mode, all the windows and doors are open. ACH with two ventilation modes, under idealized isothermal conditions or indoor heating conditions are calculated, and the results are plotted in Figure 6. Referring to the national standards, the accepted natural ventilation should have an ACH ≥ 1.5.

Traffic Pollutants Diffused from Outdoor to Indoor Environment
The parameter normalized concentration K [74] as defined in the Section 2.2.2 is applied to investigate the dispersion of the traffic-related pollutants (i.e., CO in this work) in different floors. In the real world, the emission strength in different regions usually has a huge difference, even at different orders. The investigation in terms of K is a universal method for better illustrating the deviation in different research. The influence caused by different emission strengths could be eliminated.
K at different floors in two ventilation scenarios under idealized isothermal or indoor heating conditions are calculated, and the results are plotted in Figure 7. In single-side ventilation scenarios, large K value exists in the curb-side building sides of the target canyon. In the leeward side of the target canyon, K value decreases with the rising floors. In the windward side of the target canyon, K value also decreases with the increasing floor level with isothermal conditions. When the indoor wall is heated, the K value decreases rapidly in the middle floors (Floor 3-5), with the smallest K in Floor 5. Consequently, K increases on Floors 6-7, and decreases again on the top floor. In cross-ventilation scenarios, K under isothermal condition has significant vertical variations. In Building 1, the middle floors have large K values, while the lower floors and the top floor have small K values. The vertical profile of K in the leeward side of Building 2 has an opposite variation pattern with that in Building 1, while K variation in the windward side of Building 2 is negligible. The indoor heating condition significantly decreases the indoor CO concentration due to the thermal force and the changing flow field in street canyons. The corresponding K at different floors has almost no change in all four building sides.

Quantitative Analysis of the Intake Fraction IF
The factor intake fraction IF as defined in the Section 2.2.3 is used to numerically investigate the inhalation exposure of traffic-related air pollutant (i.e., CO in this work) for the population living in the two target buildings. Assuming in each side of the two buildings, there are two families living in each building floor, and each family has 5 members. Thus, the total number of residents (P) in the two buildings are: P = 5 (members) × 2 (families) × 8 (floors) × 4 (building sides) = 320. (9) Atmosphere 2022, 13, 28

of 27
The population composition and the environmental categories employed for IF calculation is based on a study in the real urban area of Hong Kong [61]. The population composition is referred to the census data of Hong Kong in 2004 [61]. Br data is derived from the survey conducted by Allen et al. [76]. ∆t in different scenarios is obtained from telephone questionnaires to Hong Kong residents [77]. Table 2 presents the critical parametres for exposure assessment. It must be pointed out that, this work focuses on the difference of inhalation exposure between the indoor environment and outdoor environment by curb-side.

Quantitative Analysis of the Intake Fraction IF
The factor intake fraction IF as defined in the Section 2.2.3 is used to numerically investigate the inhalation exposure of traffic-related air pollutant (i.e., CO in this work) for the population living in the two target buildings. Assuming in each side of the two buildings, there are two families living in each building floor, and each family has 5 members. Thus, the total number of residents (P) in the two buildings are: P = 5 (members) × 2 (families) × 8 (floors) × 4 (building sides) = 320.
(9)  IF calculated for the four cases are plotted in Figure 8. IF of the target population ranges between 32.41-364.09 ppmv, corresponding to Case [cross, 0] and Case [singlesided, 100], respectively. Without considering the age groups, P_IF is ranges between 0.10-1.14 ppmv. In isothermal scenarios, if the ventilation mode of buildings was changed from a single-side ventilation to cross ventilation, IF would increase 1.3 times. In indoor heating scenarios, the alteration of the ventilation mode has a slight influence on IF. For the same ventilation mode, the indoor-outdoor air temperature difference produced by indoor heating could improve the dispersion of indoor air pollutants. The thermal force enhances the updraft with large velocity, contributing to dilute the air pollutants in the canyon and transport it out of the canyon. Furthermore, the ratio of indoor and outdoor IF ranges from 0.5 to 4.1, and the indoor heating has little influence on outdoor IF.
The population composition and the environmental categories employed for IF calculation is based on a study in the real urban area of Hong Kong [61]. The population composition is referred to the census data of Hong Kong in 2004 [61]. Br data is derived from the survey conducted by Allen et al. [76]. Δt in different scenarios is obtained from telephone questionnaires to Hong Kong residents [77]. Table 2 presents the critical parametres for exposure assessment. It must be pointed out that, this work focuses on the difference of inhalation exposure between the indoor environment and outdoor environment by curb-side. IF calculated for the four cases are plotted in Figure 8. IF of the target population ranges between 32.41-364.09 ppmv, corresponding to Case [cross, 0] and Case [singlesided, 100], respectively. Without considering the age groups, P_IF is ranges between 0.10-1.14 ppmv. In isothermal scenarios, if the ventilation mode of buildings was changed from a single-side ventilation to cross ventilation, IF would increase 1.3 times. In indoor heating scenarios, the alteration of the ventilation mode has a slight influence on IF. For the same ventilation mode, the indoor-outdoor air temperature difference produced by indoor heating could improve the dispersion of indoor air pollutants. The thermal force enhances the updraft with large velocity, contributing to dilute the air pollutants in the canyon and transport it out of the canyon. Furthermore, the ratio of indoor and outdoor IF ranges from 0.5 to 4.1, and the indoor heating has little influence on outdoor IF.

Exposure Concentration in the Target Street Canyon
Exposure concentration (E) defined in the Section 2.2.2 is adopted for the exposure assessment for the indoor and outdoor environment in the target canyon. As presented in Figure 9a,b, the middle floors of the leeward building side have larger E comparing to the lower and upper floors. In contrast, Middle floors of the windward building side have smaller E. With cross-ventilation, E of each floor is much smaller than that with single-side ventilation. It indicates that the cross-ventilation mode improves the indoor ventilation and the indoor air quality. Moreover, the thermal force is beneficial for diluting the indoor air pollutants. Indoor heating can decrease E in the indoor environment. Figure 9c summarizes the mean indoor and outdoor E in all four cases. Since the long duration time in the indoor environment (about 90% of the daily time), the indoor E is much higher than outdoor E, except in Case [cross, 100]. The largest value of indoor and outdoor ratio is 4.9. With isothermal conditions, the cross-ventilation mode has larger E. Furthermore, indoor heating is helpful to decrease the exposure concentration. The change of the ventilation mode only produces small E variations when indoor heating exists.

Exposure Concentration in the Target Street Canyon
Exposure concentration (E) defined in the Section 2.2.2 is adopted for the exposure assessment for the indoor and outdoor environment in the target canyon. As presented in Figure 9a, b, the middle floors of the leeward building side have larger E comparing to the lower and upper floors. In contrast, Middle floors of the windward building side have smaller E. With cross-ventilation, E of each floor is much smaller than that with singleside ventilation. It indicates that the cross-ventilation mode improves the indoor ventilation and the indoor air quality. Moreover, the thermal force is beneficial for diluting the indoor air pollutants. Indoor heating can decrease E in the indoor environment. Figure  9c summarizes the mean indoor and outdoor E in all four cases. Since the long duration time in the indoor environment (about 90% of the daily time), the indoor E is much higher than outdoor E, except in Case [cross, 100]. The largest value of indoor and outdoor ratio is 4.9. With isothermal conditions, the cross-ventilation mode has larger E. Furthermore, indoor heating is helpful to decrease the exposure concentration. The change of the ventilation mode only produces small E variations when indoor heating exists. (a)

Discussion
Our previous work studied the outdoor wind and thermal environments in a 2D idealized street canyon at a reduced canyon scale and explored the impacts of various AR as well as solar wall heating [73,78,79]. A single-vortex flow pattern was found in the street canyon at AR ~ 1 under isothermal conditions. It is similar as the flow field structure in this work with isothermal conditions (Figure 3a). The two-vortices structure is found in canyon 2 with indoor wall heating in the windward building (Figure 3c). A similar structure was also found in the street canyon at AR ~ 1 with solar wall heating (15:00) on the windward wall [73]. Previous studies focus on the outdoor air flow and the urban ventilation. This work is a further study of the previous work, focusing on the indoor-

Discussion
Our previous work studied the outdoor wind and thermal environments in a 2D idealized street canyon at a reduced canyon scale and explored the impacts of various AR as well as solar wall heating [73,78,79]. A single-vortex flow pattern was found in the street canyon at AR~1 under isothermal conditions. It is similar as the flow field structure in this work with isothermal conditions (Figure 3a). The two-vortices structure is found in canyon 2 with indoor wall heating in the windward building (Figure 3c). A similar structure was also found in the street canyon at AR~1 with solar wall heating (15:00) on the windward wall [73]. Previous studies focus on the outdoor air flow and the urban ventilation. This work is a further study of the previous work, focusing on the indoor-outdoor flow field and air mass exchange, and extending the application of CFD modelling to the area of exposure and public health. The impacts of indoor-outdoor air temperature difference on the flow, pollutants dispersion, and exposure concentration are discussed based on the numerical simulation at full scale with AR = 1. The functions of different natural ventilation modes, isothermal conditions, and indoor wall heating conditions on the indoor and outdoor flow, traffic-related pollutant dispersion, and residents' exposure are investigated. The results of the population intake fraction (IF) and personal intake fraction (P_IF) are similar with that of Hong Kong calculated by Luo et al. [61], the annual mean IF being~270 ppmv.
Altering the ventilation mode from single-side ventilation to cross ventilation under isothermal conditions would deteriorate the indoor air quality due to the diffusion of outdoor air pollutants. However, the indoor-outdoor air temperature differences could greatly improve the dispersion condition, decrease the pollutants' concentrations, and reduce indoor IF. Nevertheless, the cascading effect found in this work is an often-overlooked pathway of exposure. This effect was also found by many other previous research, both with numerical simulations [21,22] and scaled outdoor observations [29]. Furthermore, with cross ventilation, the flow from outdoor to indoor and the flow passing through the whole building could be another exposure pathway between different rooms or even different buildings, as concluded by [20]. Thus, the CFD modelling can provide a powerful tool of back trajectory for the air pollutants exposure or even for respiratory infectious diseases.
To simplify the calculation, the parameters used in this work are idealized and simplified. The idealized 2D street canyon model is employed in this work. Natural ventilation modes are adopted only for two buildings in the target street canyon. The inert gas CO is applied as the tracer gas for the traffic emission. Chemical and photochemical processes are not considered. Indoor-heating scenarios are designed for the possible ventilation scheme designed to balance energy consumption of indoor heating and indoor air quality in cold season. The heat flux setting for the indoor room wall of indoor-heating scenarios is also an idealized assumption for the cold season. Nevertheless, parameters in the real city are more complicated and are influenced by various environmental factors. One needs more consideration when applying the results to real urban communities. In ongoing work, more parameters are considered in the simulation, such as AR, outdoor solar wall heating, and urban layouts (e.g., building configurations and urban greening). Moreover, the planned work also includes model validation by our scaled outdoor experiments (H = 1.2 m) as reported by Chen et al. [78,79]. These works will be adopted in numerical studies for full-scale realistic or idealized urban models.

Conclusions
This work sets up a useful method to investigate the impacts of different building ventilation schemes and the indoor-out door air temperature differences on the flow field and the traffic-related pollutant dispersion. The application of the CFD modelling is extended to the area of exposure and consequently connected to public health, which is a novel approach. Ventilation efficiency and residents' exposure with two types of natural ventilation schemes of single-side ventilation and cross ventilation are investigated, integrating with isothermal conditions and indoor wall heating conditions, respectively. The cross ventilation improved the indoor ventilation and pollutants' dilution with isothermal conditions. The existence of the indoor-outdoor temperature difference increases the ventilation efficiency of the single-side ventilation scheme. However, it has negative impacts on the cross-ventilation mode in some floors. Moreover, the existence of indoor wall heating can clearly decrease the risk of indoor personal exposure under both natural ventilation modes. In general, cross ventilation mitigates the environmental health stress for indoor environments with indoor wall heating, and the personal intake fraction is decreased by about 66% compared to single-side ventilation. The study numerically investigates the building ventilation and pollutant dispersion in urban communities with natural ventilation.
The method and the results are helpful references for optimizing building ventilation plans and improving indoor air quality. Owing to the complex environmental in the real city, more parameters should be considered when applying the method for the real urban community. Next steps of the ongoing work are (1) coupling the radiation model in the simulation for analysing the impact of solar radiation, (2) applying the method in street canyons with various AR values, and (3) implementing scaled-model outdoor field observation campaign for further validations.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
Our previous works have implemented detailed validations for the flow and dispersion model applied in this work. Figure A1 presents the CFD validation with wind-tunnel experiment of RNG k-ε model [73]. The validation for the pollutant dispersion model is illustrated in Figure A2 [72]. investigates the building ventilation and pollutant dispersion in urban communities with natural ventilation. The method and the results are helpful references for optimizing building ventilation plans and improving indoor air quality. Owing to the complex environmental in the real city, more parameters should be considered when applying the method for the real urban community. Next steps of the ongoing work are (1) coupling the radiation model in the simulation for analysing the impact of solar radiation, (2) applying the method in street canyons with various AR values, and (3) implementing scaled-model outdoor field observation campaign for further validations.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
Our previous works have implemented detailed validations for the flow and dispersion model applied in this work. Figure A1 presents the CFD validation with windtunnel experiment of RNG k-ε model [1]. The validation for the pollutant dispersion model is illustrated in Figure A2