Contributions of Indoor and Outdoor Sources to Ozone in Residential Buildings in Nanjing

Ozone has become one of the most serious air pollutants in China in recent years. Since people spend most of their time indoors, the ozone in the indoor environment could be a major factor affecting the occupants’ health. The indoor ozone in residential buildings mainly comes from two sources: outdoor atmosphere and indoor ozone produced by electrical devices. In this study, a typical residence in Nanjing was taken as an example to calculate and compare the contributions of indoor and outdoor sources to ozone in the building. A questionnaire survey about the type, the placement, and the frequency of use of the ozone emission devices was performed to provide the basis for the settings of indoor ozone sources. The multi-zone software CONTAM was used hourly to simulate the ozone concentration in summer and in winter with inner doors either closed or open, and it was noted whether there were ozone emission devices indoors or not. Source contribution was quantified and compared by three methods in this paper: (1) the average indoor/outdoor (I/O) ratio, (2) the I/O ratio frequency, and (3) the ratio of indoor ozone concentration without ozone sources to that with ozone sources. The results showed that the contribution of outdoor sources was much greater than that of indoor sources in summer, but in winter, the frequency of I/O > 1 could reach 55.8% of the total seasonal time, and the ratio of indoor ozone concentration without sources to that with sources could reach as high as 74.3%. This meant that the indoor concentration had the potential to exceed the outdoor. Furthermore, human respiratory exposure in different ages and genders was calculated. It was found that teenagers aged 10–18 years old and female adults had a higher respiratory exposure level.


Introduction
In recent years, with the rapid development of economy and the acceleration of urbanization, air pollution in China has continued to increase. This not only influences people's health outdoors but also affects the indoor environment through windows and doors. Ozone has gradually become one of the most serious air pollutants in China, and it may cause severe photochemical pollution. The average ozone concentration in Nanjing has increased by over 25% in the last five years [1]. Numerous studies have suggested that too much exposure to ozone may cause respiratory morbidity, impaired lung function, local or systemic inflammation, cardiovascular disease, and cancer in human beings [2][3][4][5][6][7].
There are two main sources of ozone in residential buildings-the outdoor sources that enter the room mainly through natural ventilation, mechanical ventilation, and infiltration through cracks in the building envelope [8], and the indoor sources that are emitted by some electrical devices. Ground-level ozone is mainly produced by photochemical reactions of oxygen, nitrogen oxides (NOx), and volatile organic compounds (VOCs) in the atmosphere. The commonly used indoor ozone emission devices in the literature include photocopying, air purifying, disinfecting devices, etc. [9][10][11][12][13][14][15]. The ozone emission mechanisms can generally be classified into two categories: photochemical mechanisms Int. J. Environ. Res. Public Health 2019, 16, 2587 3 of 16 with a building area below 90 m 2 accounted for 58.3%, those between 90 m 2 and 144 m 2 accounted for 31.1%, and those over 144 m 2 accounted for 7.7%. The sixth national census of Nanjing showed that the number of three-person households accounted for the largest proportion, which was 35.83% [29]. Based on the above two points, we took three-person households (i.e., two-bedroom households) as an example and combined the building area range and the Nanjing housing market to determine a typical household type. It is a six-room residence in Nanjing, whose functions, areas, and relative positions of each room are shown in Figure 1. The selection of the building model could only represent part of the housing situation in Nanjing, but the research methods and the results are applicable to most scenarios.
In this paper, the multi-zone simulation software CONTAM (CONTAM 3.2, National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA) was used to create a model of the apartment. The small diamonds on the walls of the building were the openings, including the inner and the outer doors and windows. Their sizes are included in Figure 1. The floor height of residential buildings in Nanjing is generally 2.8-3 m. Except for the girder height and the ceiling height, the net height of this building model is 2.4 m.
Int. J. Environ. Res. Public Health 2019, 16, x 3 of 16 Nanjing (N 31°14'-32°37', E 118°22'-119°14'), a major city in eastern China, has a population of more than 8 million. According to the statistical yearbook by the government [28], new residential buildings in Nanjing in the past eight years were mainly apartment houses. Among these, apartments with a building area below 90 m 2 accounted for 58.3%, those between 90 m 2 and 144 m 2 accounted for 31.1%, and those over 144 m 2 accounted for 7.7%. The sixth national census of Nanjing showed that the number of three-person households accounted for the largest proportion, which was 35.83% [29]. Based on the above two points, we took three-person households (i.e., two-bedroom households) as an example and combined the building area range and the Nanjing housing market to determine a typical household type. It is a six-room residence in Nanjing, whose functions, areas, and relative positions of each room are shown in Figure 1. The selection of the building model could only represent part of the housing situation in Nanjing, but the research methods and the results are applicable to most scenarios.
In this paper, the multi-zone simulation software CONTAM (CONTAM 3.2, National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA) was used to create a model of the apartment. The small diamonds on the walls of the building were the openings, including the inner and the outer doors and windows. Their sizes are included in Figure 1. The floor height of residential buildings in Nanjing is generally 2.8-3 m. Except for the girder height and the ceiling height, the net height of this building model is 2.4 m. This study focused on the two most unfavorable situations for indoor ozone pollution, namely, opening all outer windows in summer and closing all outer windows in winter. Indoor temperatures in summer and winter were set at 26 and 18 °C, respectively, according to air conditioning design standards. Outdoor meteorological parameters in this study were the actual data from meteorological station in Nanjing in 2018. The main wind direction in summer in Nanjing is from the southeast and in winter from the northeast.
The air flow path in CONTAM software was the passageway between zones as well as between zones and atmosphere. Each zone connected to the exterior through windows and doors, while the interior connected only through doors, and the air of each room was well mixed. In this study, we assumed that the air flow in all openings was one-way flow. When doors and windows were opened, the effective ventilation area was equal to the actual opening area. When doors and windows were closed, the small area leakage model was adopted, and the effective leakage area could be calculated according to Equation (1) [30,31]. This study focused on the two most unfavorable situations for indoor ozone pollution, namely, opening all outer windows in summer and closing all outer windows in winter. Indoor temperatures in summer and winter were set at 26 and 18 • C, respectively, according to air conditioning design standards. Outdoor meteorological parameters in this study were the actual data from meteorological station in Nanjing in 2018. The main wind direction in summer in Nanjing is from the southeast and in winter from the northeast.
The air flow path in CONTAM software was the passageway between zones as well as between zones and atmosphere. Each zone connected to the exterior through windows and doors, while the interior connected only through doors, and the air of each room was well mixed. In this study, we assumed that the air flow in all openings was one-way flow. When doors and windows were opened, the effective ventilation area was equal to the actual opening area. When doors and windows were closed, the small area leakage model was adopted, and the effective leakage area could be calculated according to Equation (1) [30,31]. where L is effective leakage area (m 2 ), Q r is leakage air volume (m 3 /s), ρ is air density (m 3 /kg), ∆P r is pressure difference (Pa), C d is dimensionless flow coefficient, ∆P r , ρ and C d are usually constant and equal to 10 Pa, 1.27 m 3 /kg, and 0.6, respectively, Q r can be calculated by area index method, and 50% surplus is set aside. The size and the effective area of each opening are shown in Table 1.

Questionnaire Survey on Indoor Ozone Emission Devices
To determine the use of indoor ozone emission devices in residential buildings in Nanjing, an online questionnaire was distributed primarily to long-time Nanjing residents. A total of 316 available questionnaires were collected, each dealing with the use of ozone emission devices. The questionnaire is shown in Table 2

Contributions Quantification
In this paper, three methods were used to quantify the contribution of indoor and outdoor ozone sources. I/O was the ratio of indoor ozone concentration to outdoor concentration. It could be expressed by the Equation (2).
where c in and c out are indoor and outdoor ozone contribution. When I/O was greater than, less than, or equal to 1, it indicated that the contribution of indoor ozone sources was greater than, less than, or equal to the contribution of outdoor ozone sources. I/O frequency analysis was based on the I/O ratio. Hourly I/O values in summer and winter were calculated, and the frequency of different I/O intervals was counted. The method of judging contributions was similar to the I/O method. The higher the frequency of I/O > 1, the greater the contribution of indoor sources was.
P o was the ratio of indoor ozone concentration without indoor ozone sources to that with indoor ozone sources. It could be expressed by the Equation (3).
where CONTRI in and CONTRI out are indoor and outdoor source contributions, respectively. When there were no ozone sources indoors, it meant that the indoor ozone came entirely from outdoors. Therefore, indoor ozone concentration without indoor ozone sources represented the outdoor source contributions. When there were ozone sources indoors, the indoor ozone was from the combination of indoor and outdoor sources. Thus, P o represents the proportion of outdoor ozone contribution in the total contribution.

Questionnaire Survey on Indoor Ozone Emission Devices
The questionnaire survey showed that 68.4% of the residential buildings had ozone emission devices. Among them, 28.2% of the residences had photocopying devices, 15.4% had disinfecting devices, and 30.8% had air purifying devices, whereas some other ozone emission devices, such as refrigerators with a purification function, shoe cleaners, etc., accounted for a total of 25.6%. This paper mainly focused on the first three kinds of devices for their detailed study. Figure 2 shows the questionnaire analysis. It can be seen from Figure 2a that photocopying and air purifying devices were mainly located in the bedroom and the living room, and disinfecting devices were mainly placed in the kitchen. This was determined by the function of the devices themselves. Figure 2b shows that most residents used photocopying devices 3-4 times a week and used air purifying devices every day. For disinfecting devices, the percentage of using 1-2 times a week and every day was the same, both accounting for 40%. Figure 2c shows that photocopying devices were often used from 11:00-14:00 and from 14:00-17:00, disinfecting devices were often used from 11:00-14:00 and from 17:00-20:00, and air purifying devices were often used from 20:00-24:00. This was related to residents' living habits, including diet, sleep, and so on. By obtaining the model numbers of the devices and looking up the information of the products at the factory, ozone emission rates of various kinds of devices were analyzed, as shown in Figure 2d. The ozone emission rates of photocopying devices were 1.7-9.0 mg/h, the rates of disinfecting devices were 6.2-20.0 mg/h, and the rates of air purifying devices were 10-137 mg/h. It was obvious that air purifying devices were the main indoor ozone sources, especially those with an ozone purification function. The results of the questionnaire survey provided basis for the settings of indoor ozone sources.

Natural Ventilation without Indoor Ozone Sources
The natural ventilation under the two most unfavorable conditions with inner doors both closed and open was studied in detail. Both wind pressure and thermal pressure were considered. The summer condition was represented by three days near the summer solstice (21-23 June), and winter was represented by three days near the winter solstice (21-23 December). The hourly air exchange rate of the building average is shown in Figure 3. It can be seen that the ventilation of the building in summer was much better than that in winter, and the air exchange rate was one order of magnitude larger. It was related to the settings of the two most unfavorable conditions. The air exchange rate in summer had multiple peaks in a day, while in winter, there was usually only one peak in a day, and it occurred around 12:00. The opening of the inner doors in summer could improve the ventilation, but in winter, the ventilation had little to do with the inner doors.

Natural Ventilation without Indoor Ozone Sources
The natural ventilation under the two most unfavorable conditions with inner doors both closed and open was studied in detail. Both wind pressure and thermal pressure were considered. The summer condition was represented by three days near the summer solstice (21-23 June), and winter was represented by three days near the winter solstice (21-23 December). The hourly air exchange rate of the building average is shown in Figure 3. It can be seen that the ventilation of the building in summer was much better than that in winter, and the air exchange rate was one order of magnitude larger. It was related to the settings of the two most unfavorable conditions. The air exchange rate in summer had multiple peaks in a day, while in winter, there was usually only one peak in a day, and it occurred around 12:00. The opening of the inner doors in summer could improve the ventilation, but in winter, the ventilation had little to do with the inner doors.

Natural Ventilation without Indoor Ozone Sources
The natural ventilation under the two most unfavorable conditions with inner doors both closed and open was studied in detail. Both wind pressure and thermal pressure were considered. The summer condition was represented by three days near the summer solstice (21-23 June), and winter was represented by three days near the winter solstice (21-23 December). The hourly air exchange rate of the building average is shown in Figure 3. It can be seen that the ventilation of the building in summer was much better than that in winter, and the air exchange rate was one order of magnitude larger. It was related to the settings of the two most unfavorable conditions. The air exchange rate in summer had multiple peaks in a day, while in winter, there was usually only one peak in a day, and it occurred around 12:00. The opening of the inner doors in summer could improve the ventilation, but in winter, the ventilation had little to do with the inner doors.    Table 3. It was found that Bedroom A had the best ventilation; the air exchange rate in summer with inner doors opening could reach as high as 78.58 h −1 . Ozone concentration also varied with the season and the state of the inner doors. When the inner doors were open in summer, the ozone concentration of all rooms was higher than that when the inner doors were closed, except Bedroom A, since Bedroom A was on the windward side of the building. When the inner doors were open in winter, the concentration of each room increased slightly. When there were no indoor sources, the indoor ozone came entirely from the outdoors. The natural ventilation pattern without indoor ozone sources served as the basis for calculating the contributions of the sources in the later section.
The effects of different factors on indoor ozone concentration under natural ventilation were also studied, including temperature, wind speed, wind direction, and ozone deposition velocity. Table 4 shows the settings of influencing factors. In this paper, the variable-controlling method was used to study the effects of different influencing factors on indoor ozone concentration. The first row shows the average value of different factors in summer and winter. Using temperature as an example, 25% and 75% percentiles were used to study the effect of temperature on indoor ozone concentration in the living room, the kitchen, and Bedroom A, while other factors remained unchanged. The same was true of other influencing factors. The results are shown in Figure 4. It can be seen that when the inner doors were closed in summer, decreasing the outdoor temperature from 28 • C to 26 • C reduced the ozone concentration of Bedroom A by 28.3%. In winter, the ozone concentration in the living room was more sensitive to temperature changes. The ozone concentration in the kitchen showed minimal effect by temperature, and it showed a great effect from wind speed, wind direction, and deposition. When inner doors were closed in summer, increasing wind speed, decreasing wind angle, and deposition velocity could lead to a great ozone variation in the kitchen. As for ozone deposition velocity, there was more room for indoor ozone concentration to decrease. Generally speaking, indoor ozone concentration in winter was affected more by the above factors than in summer. It can be seen that when the inner doors were closed in summer, decreasing the outdoor temperature from 28 ℃ to 26 ℃ reduced the ozone concentration of Bedroom A by 28.3%. In winter, the ozone concentration in the living room was more sensitive to temperature changes. The ozone concentration in the kitchen showed minimal effect by temperature, and it showed a great effect from wind speed, wind direction, and deposition. When inner doors were closed in summer, increasing wind speed, decreasing wind angle, and deposition velocity could lead to a great ozone variation in the kitchen. As for ozone deposition velocity, there was more room for indoor ozone concentration to decrease. Generally speaking, indoor ozone concentration in winter was affected more by the above factors than in summer.

Indoor Ozone Level with Ozone Emission Devices.
This section mainly studied the ozone concentration level of each room when there were ozone emission devices in the building. According to the questionnaire survey, we used CONTAM to simulate the indoor ozone concentration hourly when photocopying devices were located in Bedroom A and the living room, when air purifying devices were in Bedroom A and the living room, and when disinfecting devices were in the kitchen. The photocopying devices worked from 11:00 to 14:00 on Monday and Wednesday and from 14:00 to 17:00 on Friday and Sunday. The median ozone emission rate 5 mg/h was adopted. The air purifying devices worked from 20:00 to 24:00 every day, and the median ozone emission rate 38.8 mg/h was adopted. The disinfecting devices worked from 11:00 to 14:00 and from 17:00 to 20:00 on Friday and Saturday, and the median ozone emission rate 14.4 mg/h was adopted. The simulation time step was 5 min, and the total duration was 3 months in summer or winter. Whether the inner doors were closed or open was also considered in this section.
Taking the air purifying devices as an example, this paper studied the variation of the ozone concentration level of the source room and the adjacent representative room when ozone was emitted, as shown in Figure 5.

Indoor Ozone Level with Ozone Emission Devices.
This section mainly studied the ozone concentration level of each room when there were ozone emission devices in the building. According to the questionnaire survey, we used CONTAM to simulate the indoor ozone concentration hourly when photocopying devices were located in Bedroom A and the living room, when air purifying devices were in Bedroom A and the living room, and when disinfecting devices were in the kitchen. The photocopying devices worked from 11:00 to 14:00 on Monday and Wednesday and from 14:00 to 17:00 on Friday and Sunday. The median ozone emission rate 5 mg/h was adopted. The air purifying devices worked from 20:00 to 24:00 every day, and the median ozone emission rate 38.8 mg/h was adopted. The disinfecting devices worked from 11:00 to 14:00 and from 17:00 to 20:00 on Friday and Saturday, and the median ozone emission rate 14.4 mg/h was adopted. The simulation time step was 5 min, and the total duration was 3 months in summer or winter. Whether the inner doors were closed or open was also considered in this section.
Taking the air purifying devices as an example, this paper studied the variation of the ozone concentration level of the source room and the adjacent representative room when ozone was emitted, as shown in Figure 5.  Figure 5 shows the hourly variation of ozone concentration in Bedroom A and the living room on a day when indoor sources were located in one of them. When the ozone emission devices started working at 20:00 hours, the concentration of the source room would have a significant and sudden increase, while the other room would change slightly or would remain unchanged. This was because ozone could spread to other rooms through inner doors or cracks, but when the source room was located in the downwind direction, it had less impact on other rooms, just like the case where the sources were located in the living room [LR(SIBRA)] in Figure 5(a) and (b). Taking Figure 5(a) as an example, when inner doors were closed in summer and the source was located in Bedroom A, it took about 30 minutes for Bedroom A to reach the maximum 85 ppb, which was consistent with the results of literature [32]. After the devices were shut down, its concentration returned to the background concentration rapidly. The ozone concentration in Bedroom A exceeded 50 ppb for three hours, thus threatening the health of residents. In winter, the ozone concentration in source rooms was particularly high, indicating a need for additional attention. Figure 5 shows the ozone level for only one day when air purifying devices were located in the building. Different types of devices and different source locations were also simulated hourly. The average value is shown in Table 5. It can be seen that when inner doors were open in summer, the ozone level was nearly two times higher than that with inner doors closed, except in Bedroom A. In winter, the state of inner doors had little effect.   Figure 5 shows the hourly variation of ozone concentration in Bedroom A and the living room on a day when indoor sources were located in one of them. When the ozone emission devices started working at 20:00 h, the concentration of the source room would have a significant and sudden increase, while the other room would change slightly or would remain unchanged. This was because ozone could spread to other rooms through inner doors or cracks, but when the source room was located in the downwind direction, it had less impact on other rooms, just like the case where the sources were located in the living room [LR(SIBRA)] in Figure 5a,b. Taking Figure 5a as an example, when inner doors were closed in summer and the source was located in Bedroom A, it took about 30 min for Bedroom A to reach the maximum 85 ppb, which was consistent with the results of literature [32]. After the devices were shut down, its concentration returned to the background concentration rapidly. The ozone concentration in Bedroom A exceeded 50 ppb for three hours, thus threatening the health of residents. In winter, the ozone concentration in source rooms was particularly high, indicating a need for additional attention. Figure 5 shows the ozone level for only one day when air purifying devices were located in the building. Different types of devices and different source locations were also simulated hourly. The average value is shown in Table 5. It can be seen that when inner doors were open in summer, the ozone level was nearly two times higher than that with inner doors closed, except in Bedroom A. In winter, the state of inner doors had little effect. When there were ozone emission devices in the building, the indoor ozone was from the combination of indoor and outdoor sources. The conditions with indoor ozone sources would also be the basis for the calculation of the contributions of the sources.

Indoor and Outdoor Ozone Contributions
In this paper, three methods were used to quantify and compare the contributions of indoor and outdoor ozone sources. The most commonly used method is to calculate the average I/O ratio of the building. According to the simulation in previous sections, average indoor and outdoor concentrations of the building could be obtained. The average I/O ratios of the main rooms are shown in Table 6. When there were no ozone sources in the building, the I/O ratio of the living room, Bedroom A, and the kitchen reached more than 0.7 when inner doors were open in summer. However, in winter, no matter whether the inner doors were closed or open, the I/O ratio of each room was small. The I/O ratio of Bedroom A decreased when inner doors were open, whereas other rooms increased. This meant that Bedroom A was most disadvantageous to ozone spread. When there were ozone emission devices in the building, the ozone level of all scenarios had a certain increase, especially in winter. The ozone level of the source room was greater than that of the building average, and this was due to the fact that the indoor ozone in other rooms was all spread from the source room through the inner doors or the cracks. As for different ozone emission devices, they contributed almost the same to indoor ozone in summer, but in winter, air purifying devices contributed the most of the three kinds of devices. Only when the air purifying devices were located in Bedroom A in winter was the I/O ratio greater than 1, meaning that indoor sources contributed more than outdoor sources. In this case, residents should control the use of ozone emission devices and open windows for reducing ozone exposure. In addition to such a condition, when I/O was less than 1, the contribution of outdoor sources was greater than that of indoor sources.
Average I/O ratio could only roughly show the indoor and the outdoor ozone concentration, since it was affected by many other factors. Therefore, in order to make a more detailed analysis, the ozone I/O ratios of Bedroom A and the living room in summer (1 June-31 August) and winter (1 December-28 February) were calculated hourly, and the frequency of I/O ratio in different ranges is shown in Figure 6.
The darker bar indicates that the I/O ratio was greater than 1, i.e., the concentration indoors was higher than that outdoors. In summer, no matter which kind of devices or in which room the ozone emission devices were located, the frequencies of I/O > 1 of both the source room and the building average were no more than 13%, indicating that outdoor sources were absolutely dominant. Among the three kinds of devices, air purifying devices had the highest frequency of IO > 1. Moreover, the difference between scenarios of the same devices was not significant for I/O > 1. In winter, the frequency of I/O > 1 was higher than that in summer, accounting for 7.5-55.8%, but only when air purifying devices were located in Bedroom A could the frequency of I/O > 1 exceed 50%, meaning that indoor sources became dominant. The result was the same with the method of calculating the average I/O ratio, but the I/O frequency contained more information.
In addition to the analysis of I/O, the ratio of concentrations in the building without and with ozone sources could also be used to quantify and compare the contributions of indoor and outdoor sources. Tables 3 and 5  windows for reducing ozone exposure. In addition to such a condition, when I/O was less than 1, the contribution of outdoor sources was greater than that of indoor sources. Average I/O ratio could only roughly show the indoor and the outdoor ozone concentration, since it was affected by many other factors. Therefore, in order to make a more detailed analysis, the ozone I/O ratios of Bedroom A and the living room in summer (1 June-31 August) and winter (1 December-28 February) were calculated hourly, and the frequency of I/O ratio in different ranges is shown in Figure 6.
The darker bar indicates that the I/O ratio was greater than 1, i.e., the concentration indoors was higher than that outdoors. In summer, no matter which kind of devices or in which room the ozone emission devices were located, the frequencies of I/O > 1 of both the source room and the building average were no more than 13%, indicating that outdoor sources were absolutely dominant. Among the three kinds of devices, air purifying devices had the highest frequency of IO > 1. Moreover, the difference between scenarios of the same devices was not significant for I/O > 1. In winter, the frequency of I/O > 1 was higher than that in summer, accounting for 7.5-55.8%, but only when air purifying devices were located in Bedroom A could the frequency of I/O > 1 exceed 50%, meaning that indoor sources became dominant. The result was the same with the method of calculating the average I/O ratio, but the I/O frequency contained more information.
In addition to the analysis of I/O, the ratio of concentrations in the building without and with ozone sources could also be used to quantify and compare the contributions of indoor and outdoor sources. Table 3 and Table 5 calculate the ozone concentrations of the main rooms without and with ozone sources. By dividing the concentration of the corresponding rooms, the contributions of both indoor and outdoor sources could be calculated, as shown in Figure 7.  As can be seen in Figure 7, in summer, no matter which kind of ozone emission devices or in which room the indoor sources were located, the contributions of outdoor sources were absolutely dominant, all exceeding 70%. The highest indoor source contribution also occurred in air purifying devices. When inner doors were closed, the contributions of indoor sources were greater on the same windows for reducing ozone exposure. In addition to such a condition, when I/O was less than 1, the contribution of outdoor sources was greater than that of indoor sources. Average I/O ratio could only roughly show the indoor and the outdoor ozone concentration, since it was affected by many other factors. Therefore, in order to make a more detailed analysis, the ozone I/O ratios of Bedroom A and the living room in summer (1 June-31 August) and winter (1 December-28 February) were calculated hourly, and the frequency of I/O ratio in different ranges is shown in Figure 6.
The darker bar indicates that the I/O ratio was greater than 1, i.e., the concentration indoors was higher than that outdoors. In summer, no matter which kind of devices or in which room the ozone emission devices were located, the frequencies of I/O > 1 of both the source room and the building average were no more than 13%, indicating that outdoor sources were absolutely dominant. Among the three kinds of devices, air purifying devices had the highest frequency of IO > 1. Moreover, the difference between scenarios of the same devices was not significant for I/O > 1. In winter, the frequency of I/O > 1 was higher than that in summer, accounting for 7.5-55.8%, but only when air purifying devices were located in Bedroom A could the frequency of I/O > 1 exceed 50%, meaning that indoor sources became dominant. The result was the same with the method of calculating the average I/O ratio, but the I/O frequency contained more information.
In addition to the analysis of I/O, the ratio of concentrations in the building without and with ozone sources could also be used to quantify and compare the contributions of indoor and outdoor sources. Table 3 and Table 5 calculate the ozone concentrations of the main rooms without and with ozone sources. By dividing the concentration of the corresponding rooms, the contributions of both indoor and outdoor sources could be calculated, as shown in Figure 7.  As can be seen in Figure 7, in summer, no matter which kind of ozone emission devices or in which room the indoor sources were located, the contributions of outdoor sources were absolutely dominant, all exceeding 70%. The highest indoor source contribution also occurred in air purifying devices. When inner doors were closed, the contributions of indoor sources were greater on the same As can be seen in Figure 7, in summer, no matter which kind of ozone emission devices or in which room the indoor sources were located, the contributions of outdoor sources were absolutely dominant, all exceeding 70%. The highest indoor source contribution also occurred in air purifying devices. When inner doors were closed, the contributions of indoor sources were greater on the same condition. In winter, the contributions of indoor sources had an increase, with most contributions coming from photocopying devices at 30.9%, disinfecting devices at 40.5%, and air purifying devices at more than 50%, showing that indoor sources were dominant.
Comparing the conclusions of Figures 6 and 7, outdoor sources dominated in summer in both methods. In winter, only when air purifying devices were located in Bedroom A could the frequency of I/O > 1 exceed 50%, while the contribution of indoor sources could exceed that of outdoor sources in both Bedroom A and the living room, according to Figure 7. This was because I/O frequencies were calculated by counting the hours' proportion of different I/O ranges, but there might have been a very large I/O in a short period of time, which may have led to a high average concentration but only a little improvement on I/O > 1 time. Both of these results make sense in quantification and comparison of indoor and outdoor ozone contributions.

Human Respiratory Exposure
The total exposure of the human body to air pollutants refers to the product of pollutant concentration and the residence time when a person stays in a specific environment. Indoor air pollutant exposure could be calculated by the Equation (4) [33][34][35]: where BE is total exposure (mg), c is exposure concentration (mg/m 3 ), and t 1 and t 2 are exposure start and stop time(s), respectively. Total exposure includes respiratory exposure and skin exposure. This paper only considered respiratory exposure. There is a big difference in respiratory rate under different conditions, thus the effect of the respiratory rate on exposure should be considered. Long-term respiratory volume could be calculated by Equation (5) [36][37][38]: where IR L is long-term respiratory volume (m 3 /d), BMR is basal metabolic rate (kJ/d), E is oxygen consumption by unit energy metabolic (giving us 0.05 L/kJ), VQ is the dimensionless ventilation equivalent of 27 here, and A is the coefficient of long-term respiratory volume. BMR and A of different sex and age groups can be found in technical specifications according to height and weight. In this paper, the condition that air purifying devices were located in the building was used to study the ozone respiratory exposure. Figure 8 shows the respiratory exposure in different ages and genders. It was found that teenagers aged 10-18 years old had the highest respiratory exposure. This was because teenagers had a higher respiratory coefficient, even though they were smaller in weight. Considering that teenagers are more sensitive to air pollutants, relevant control strategies should pay more attention to their protection. The ozone exposure of adult females was higher than that of adult males. Although males had a higher long-term respiratory rate, their indoor activity time was shorter than that of females. teenagers aged 10-18 years old had the highest respiratory exposure. This was because teenagers had a higher respiratory coefficient, even though they were smaller in weight. Considering that teenagers are more sensitive to air pollutants, relevant control strategies should pay more attention to their protection. The ozone exposure of adult females was higher than that of adult males. Although males had a higher long-term respiratory rate, their indoor activity time was shorter than that of females.

Implications and Limitations of the Study
In this study, three methods were used to quantitatively estimate the contributions of indoor and outdoor sources to indoor ozone in residential buildings. Although indoor ozone levels vary with seasons, with inner doors open or closed, with meteorological parameters, as well as with different settings of ozone emission devices, indoor source contributions to indoor ozone cannot be ignored for most scenarios, especially in winter. Some studies simply utilized ambient ozone concentrations to evaluate human exposure [39,40], which may not reflect the real exposure concentration according to the results of this paper. As to the implications of this study, the results of the model for a typical residence in Nanjing revealed the significance of indoor ozone contributions, thus calling attention to the need for increased scrutiny of indoor sources of ozone. In addition, different control strategies such as closing or opening doors and windows in a timely manner or using ozone emission devices can be taken in different scenarios to minimize the human exposure to ozone and to provide for a healthier building environment.
However, there are some limitations of this research and the model implications at present. For example, the building model used in this paper represents only the most typical situation in Nanjing. Since there are so many types of residential buildings in the city, they all cannot be fully covered in this paper. However, the research methods and the results can be generalized in this region. Additionally, as a pure software simulation research, if there were real experiments for comparison, the results would be more credible and scientific. The next step of this study is to select some typical residential buildings in Nanjing to monitor the ozone concentration over a longer period of time and to record the residents' customary practices of opening and closing windows and doors so as to provide experimental support for this research. In addition to indoor and outdoor ozone sources, indoor ozone concentration is also affected by ozone removal rate. The indoor ozone concentration can be defined by Equation (6) [15].
where C in and C out are the indoor and the outdoor ozone concentration (ppb), k is the ozone penetration factor (0-1), λ is the air exchange rate (h-1), RE is the ozone emission rate of the indoor ozone devices (ppb/h), RC is the ozone removal rate by gaseous chemicals through chemical reactions (ppb/h), RM is the ozone removal rate by indoor building materials (ppb/h), and RH is the rate at which ozone is removed by human surfaces (ppb/h). The ozone removal rate on the surface of the building material and the human body was kept constant in this paper, but the chemical reaction between ozone and other gases was not considered, because it is too complex and the amount of reaction is small. Future studies should investigate the impact of different kinds of ozone removal in residential buildings in detail.

Conclusions
This study aimed to investigate the contributions of indoor and outdoor sources to ozone in residential buildings. The multi-zone software CONTAM was used to simulate the air exchange rate and the ozone concentration in a typical residence in Nanjing, China. Firstly, a questionnaire survey about the ozone emission devices was performed in Nanjing. The devices type, the placement, and the frequency of use analyzed by the questionnaire survey provided the basis for the settings of indoor ozone sources. Then, the natural ventilation pattern of the typical building model was simulated, and air exchange rates and ozone concentrations of the main rooms were obtained. The natural ventilation condition with no indoor sources served as the basis of the calculation of source contributions. The effects of different factors on indoor ozone concentration under natural ventilation were also studied, including temperature, wind speed, wind direction, and ozone deposition velocity. Next, the condition under which there were ozone emission devices indoors was simulated. Taking the air purifying devices as an example, this paper studied the variations of ozone concentration levels of the source room and the adjacent representative room when ozone was emitted. Meanwhile, the average ozone concentrations in the source room and the building average with different ozone emission devices were calculated.
Three methods were used to quantify the contribution of indoor and outdoor ozone sources in this paper, namely, the average I/O ratio, the frequency of I/O, and the ratio of indoor ozone concentration without ozone sources to that with ozone sources. According to average I/O ratio, only when air purifying devices were located in Bedroom A in winter could the indoor source contribution exceed the outdoor. I/O frequency was the proportion of the time when I/O was greater than 1, less than 0.5, and between 0.5 and 1 in the whole season. It was found by I/O frequency analysis that outdoor sources were dominant in summer, but when air purifying devices were located in Bedroom A in winter, the frequency of I/O > 1 could exceed 50%, which meant that indoor sources became dominant. The ratio of indoor ozone concentration without ozone sources to that with ozone sources was the contribution of outdoor sources. The results showed that the contributions of outdoor sources were absolutely dominant in summer, but in winter, the indoor contributions of air purifying devices were more than 50% regardless of the scenarios, meaning that the indoor sources were dominant.
Finally, respiratory exposure in different ages and genders was calculated. Teenagers aged 10-18 had the highest respiratory exposure because of the high respiratory coefficient, and the exposure of female adults was higher than males because they spent more time indoors. The results of the model for a typical residence in Nanjing revealed the significance of indoor ozone contributions, thus calling attention to the need for increased scrutiny of indoor sources of ozone. In addition, appropriate measures should be taken to minimize the human exposure to ozone both indoors and outdoors.
Author Contributions: Y.H. collated the data, wrote the first paper draft, and revised the manuscript and the protocol. Z.Y. collaborated on the paper's first draft and was responsible for the CONTAM simulation. Z.G. proposed the protocol framework, worked extensively on the manuscript and served as corresponding author.