Effect of Urbanization on Ozone and Resultant Health Effects in the Pearl River Delta Region of China
Abstract
The United Nations has reported that 55% of the global population resides in urban areas, and 68% of the population is expected to be urban dwellers by 2050. Urbanization has critical implications for global land cover. Relevant literature has provided evidence attributing climatic effects to urban expansion; however, few studies have investigated the effect on public health and pollutant sensitivity to emissions. This study aimed to characterize the effect of urbanization-induced changes in regional climate on ozone (O3), to evaluate ozone sensitivity to nitrogen oxide (NOx) and volatile organic compound (VOC) emissions, and to estimate premature mortalities due to O3 exposure. We employed atmospheric models with the higher-order decoupled direct method to simulate effects of urbanization on O3 and to determine O3 sensitivity to NOx and VOC emissions. China-specific concentration response functions were utilized to estimate cardiovascular and respiratory mortalities due to ozone exposure. Urbanization increased O3, which translated to a 39.6% increase in O3-induced premature mortality (1,100 deaths). Moreover, O3 has become less/more sensitive to unit changes in NOx and VOC emissions in various cities. Urban greening may reduce urban temperature, but it may increase O3 in some cities due to the additional VOC emissions of greening. These findings highlight the strong interactions between land use policies, urban climate adaptation strategies, and air quality policies, suggesting the need of cobeneficial strategies and policies. We proposed a precision environmental management concept that emphasizes the importance of considering the specific atmospheric condition and composition of a city when formulating its environmental policies.
Plain Language Summary
The United Nations has reported that 55% of the global population resides in urban areas, and 68% of the population is expected to be urban dwellers by 2050. Urbanization has critical implications for global land cover. Relevant literature has provided evidence attributing climatic effects to urban expansion; however, few studies have investigated the effect on public health and pollutant sensitivity to emissions. This study aimed to characterize the effect of urbanization-induced changes in regional climate on ozone (O3) and to estimate premature mortalities due to O3 exposure. Urbanization increased O3, which translated to a 39.6% increase in O3-induced premature mortality (1,100 deaths). Moreover, O3 has become less/more sensitive to unit changes in NOx and VOC emissions in various cities. Urban greening may reduce urban temperature, but it may increase O3 in some cities as a result of the VOC emissions of greening. These findings highlight the strong interactions between land use policies, urban climate adaptation strategies, and air quality policies, suggesting the need of cobeneficial strategies and policies. We proposed a precision environmental management concept that emphasizes the importance of considering the specific atmospheric condition and composition of a city when formulating its environmental policies.
1 Introduction
The United Nations has reported that 55% of the global population resides in urban areas and projects that 68% of the population will be urban dwellers by 2050 (United Nations, 2018). Urbanization has critical implications for both current and future land cover; it affects climate and air quality and, consequently, public health. To develop a sustainable urban environment, the urbanization effect should be comprehensively evaluated and quantified, especially from the perspective of public health.
Even though urbanization has been proven to have a notable effect on urban climate (Cao et al., 2016; Chen et al., 2014; Chen et al., 2015; Cheng & Chan, 2012; Huff & Changnon, 1972; Inoue & Kimura, 2004; Jingyong et al., 2005; Kalnay & Cai, 2003; Kishtawal et al., 2010; Li et al., 2014, 2015, 2016; Liao et al., 2015; Lin et al., 2007; Lo et al., 2007; Niyogi et al., 2011; Shem & Shepherd, 2009; Shi et al., 2008; Sun et al., 2016; Tayanc & Toros, 1997; Wang et al., 2007; Wang et al., 2019; Zhang et al., 2011; Zhou et al., 2004), the resultant annual and seasonal impacts on pollutant levels and atmospheric response to emissions have yet to be fully understood. A study conducted simulations to assess the urbanization effect on climate and air quality over eastern China in July from 2008 to 2012. The authors found that carbon monoxide, elemental carbon, and particulate matter with an aerodynamic diameter ≤2.5 μm (PM2.5) decreased near the ground level but increased at 1–3 km above the ground, whereas ozone increased up to 4 km above the ground (Tao et al., 2015). They also found that a 10% increase in urban land coverage causes a 1% increase in O3 but a 2% decrease in other three pollutants. Another study estimated that urban warming decreased the ozone (O3) concentration by 1.3 ppb during the daytime but increased the O3 concentration by 5.2 ppb during the nighttime in another developing region, namely, the Pearl River Delta (PRD) region (Li et al., 2016). These changes in O3 were caused by increases in wind speed and the atmospheric boundary layer depth, which led to the dilution of nitrogen oxides (NOx). The dilution then increased nocturnal O3 by suppressing NOx titration destruction and decreased daytime O3 by weakening photochemical production. A previous study evaluated the urbanization effect on the formation of secondary organic aerosols in the PRD region in March 2001 (Wang et al., 2009) and estimated that urbanization caused a reduction in NOx and volatile organic compound (VOC) concentrations by 4 and 1.5 ppb, respectively. The increase in temperature combined with a reduction in wind speed caused an increase in O3 and nitrate radical concentrations by 2–4 ppb and 4–12 ppt, respectively. These results highlight the substantial urbanization effect on monthly O3. To further understand the urbanization effect on air quality, the induced-changes in air quality over an entire year and in various months and changes in atmospheric responses to emissions must be comprehensively assessed and quantified.
While air quality is anticipated to change due to urbanization, the resultant health impact has not assessed. Relevant studies have reported that poor air quality adversely affects human health (Y. Chen et al., 2013; Guo et al., 2013; Jerrett et al., 2009; Turner et al., 2016). The Harvard Six Cities and American Cancer Society epidemiological studies have revealed that exposure to inhaled O3 may cause oxidative damage and respiratory diseases (Jerrett et al., 2009). Although the urbanization effect on air quality has been investigated in relevant literature, less attention has been paid to the consequential effects on public health. These effects therefore need to be fully understood.
This study comprehensively investigated and quantified the annual, monthly, and diurnal effects of urbanization on air quality and evaluated O3 sensitivity to NOx and VOC emissions. The consequent impacts on public health in the entire region and in each city were quantified in terms of premature mortality caused by O3 exposure. Our results may provide a quintessential resource for atmospheric scientists to understand the chemical reactions occurring as a result of urban warming and may help policy makers establish effective cobenefit policies to mitigate regional climate and air pollution problems.
This paper is structured as follows: The data and methods are described in detail in section 2. The initial results provided in section 3 present the urbanization effect on regional and urban air quality in terms of changes in the O3 concentration. We discuss the baseline (year 2010) O3 sensitivity to NOx and VOC emissions and then describe our findings in relation to the urbanization effect on O3 sensitivity to NOx and VOC emissions. The last part of the results section focuses on O3-induced premature mortality. A discussion and conclusion are provided in section 4.
2 Materials and Methods
2.1 Experimental Designs
The PRD region was selected as the study site because it has undergone rapid urbanization in the recent 30 years (Cao et al., 2016; Chen et al., 2014; Chen et al., 2015; Cheng & Chan, 2012; Li et al., 2016; Wang et al., 2007; Wang et al., 2009; Zhang et al., 2011). In this study, we employed the Weather Research and Forecast (WRF) model/Community Multiscale Air Quality Modeling System and the high-order decoupled direct method (CMAQ-HDDM; Cohan et al., 2005; Hakami et al., 2003) to simulate the effects of urbanization on O3 and determine O3 sensitivity to NOx and VOC emissions. Similar to the scenarios evaluated in other studies (Li et al., 2016; Wang et al., 2007; Wang et al., 2009), we evaluated two scenarios related to land cover: (1) lightly urbanized and (2) heavily urbanized (see Figure 1 and Table 1). The land cover data for the two scenarios were obtained from the satellite observational study by Liu et al. (2014). In the heavily urbanized scenario, 2010 served as the baseline year, and the meteorology and land cover data for 2010 were used, whereas in the lightly urbanized scenario, the meteorology for 2010 and land cover data for 1990 (prior to heavy urban expansion) were used. The two years were selected due to the highest urbanization rate in the region during the period (Chen et al., 2013). Differences in model results between the two land cover scenarios were attributed to the effect of rapid urbanization. The emissions between the two land cover scenarios were kept constant to isolate the effect of urbanization on O3.
| Scenario number | Scenario name | Land cover data | Year | IC/BC | Description |
|---|---|---|---|---|---|
| 1 | Heavily urbanized scenario | 2010 | 2010 | FNL | Simulation for 2010 using the land cover data from 2010 and driven by 2010 FNL data |
| 2 | Lightly urbanized scenario | 1990 | 2010 | FNL | Simulation for 2010 using the land cover data from 1990 and driven by 2010 FNL data to examine the effect of a lightly urbanized scenario |
- Note. FNL refers to the National Centers for Environmental Prediction/Final reanalysis data (https://rda.ucar.edu/datasets/ds083.2/). The urbanization effect was estimated from the difference between the lightly urbanized and heavily urbanized scenarios using land cover data for 2010 and 1990, respectively (Liu et al., 2014).
2.2 Meteorological Model and Its Evaluation
We employed the WRF model (v3.7; Skamarock et al., 2008) with the single-layer urban canopy model (Kusaka & Kimura, 2004) to provide meteorological data for our chemical and transport model. The WRF model was configured to possess three nested domains (Figure 2). The outermost domain (D1), with a spatial resolution of 27 km, covered the entire China and provided boundary conditions for a nested domain (D2). The nested domain, with a spatial resolution of 9 km, encompassed southern China. The innermost domain (D3), with a spatial resolution of 3 km, covered the PRD region. The model was configured to have 27 vertical sigma levels unequally spaced from the ground to the top of the model (50 hPa), with the first 17 layers being concentrated in the first kilometer above ground level to resolve the detailed structure of the planetary boundary layer (PBL). The WRF model was driven by Final Operational Global Analysis data at 6-hr temporal resolution and 1° × 1° spatial resolution (National Centers for Environmental Prediction et al., 2000). The model was evaluated using a group of measurements available in the region.
The study period was the entire year of 2010, while an additional 2-week period was also simulated to provide a spin-up period for the chemistry transport model. The 1-year WRF simulation results were staggered by multiple reinitialized WRF model runs, in each of which 96-hr meteorology was simulated with the first 24 simulated hours as a spin-up period. We note that the outputs of D2 domain were used to provide boundary conditions for D3; the land use data for various land use scenarios were used in D3 only. Further details of the model configurations and the evaluation results are provided in the Supporting Information (SI).
2.3 Chemistry Transport Model and Its Evaluation
(1)Similar to Yim et al. (2019), the initial and boundary air quality conditions were provided by GEOS-Chem v8.3.2, with a spatial resolution of 4° × 5° of GEOS-Chem (Bey et al., 2001). GEOS-Chem is a global chemical and transport model that has been commonly used in various studies on global air quality and extensively evaluated through observation (Yim et al., 2015). Emissions outside the PRD region were provided by the INTEX-B 2006 regional emission inventories (Zhang et al., 2009). The 2010 PRD emissions were obtained from the study by Yin et al. (2015) and Zheng et al. (2009), and the Hong Kong Environmental Protection Department (https://www.epd.gov.hk/epd/sites/default/files/epd/english/environmentinhk/air/data/files/2010HKEIReportEng.pdf). MEGAN v2.04 (Guenther et al., 2006) and the leaf area index derived from the Moderate Resolution Imaging Spectroradiometer were used to estimate biogenic emissions.
The CMAQ model was comprehensively evaluated using all the publicly available O3 measurements obtained from the Pearl River Delta Regional Air Quality Monitoring Network (https://www.epd.gov.hk/epd/sites/default/files/epd/english/resources_pub/publications/files/PRD_2010_report_en.pdf) and the Hong Kong Environmental Protection Department. Statistical parameters including means, standard deviation, normalized mean bias (NMB), and index of agreement were computed to quantify model performance. The evaluation results are provided in Figure S1 and Table S6 in the SI. On average, the index of agreement and NMB of O3 were approximately 0.54 (minimum: 0.37; maximum: 0.64) and 0.54% (minimum: −0.13%; maximum: 1.01%), respectively. Overall, model performance was consistent with that reported by other studies (Kwok et al., 2012; Wang et al., 2010; Wu et al., 2012) and was sufficient for this study.
2.4 Assessment of the Effect of Urbanization on Public Health
(2)
(3)2.5 Uncertainty Quantification
Our uncertainty quantification considered various uncertainties, including model performance and CRF uncertainties induced when determining the experimental coefficient for the RR calculation. Uncertainty in model performance was quantified using the average NMB estimated in the model evaluation. The statistical results revealed that under a 95% confidence interval (CI), lower bound, mean, and upper bound NMB distributions across various observations stations were −4.3%, 9.1%, and 15.1%, respectively, in the baseline year 2010 (heavily urbanized scenario). Uncertainty in cardiovascular and respiratory mortalities caused by ozone pollution was estimated in terms of the fluctuation of CRF coefficients (X0 and γ). Specifically, for cardiovascular and respiratory mortalities, X0 varied among [1.0, 16.2, 32.6] and [6.2, 29.8, 40.0], respectively, whereas γ varied among [5.4 × 10−4, 1.3 × 10−3, 2.3 × 10−3] and [4.5 × 10−4, 7.2 × 10−4, 1.2 × 10−3], respectively. An uncertainty distribution was estimated and a triangular distribution of estimates was constructed on the basis of a 2,000 member Monte Carlo simulation, in which realizations were randomly selected for each respective uncertainty factor in the calculation. The ultimate uncertainty was estimated based on the mean and 95% CI of the effect distribution.
3 Results
3.1 Effect of Rapid Urban Expansion on Air Quality
Figure 3 depicts the changes in annual average ozone concentration between the lightly urbanized and heavily urbanized land cover scenarios. In general, urbanization caused an increase in annual average regional O3 concentration, with a maximum concentration of 6.99 ppb, whereas annual average O3 concentration in urban areas increased by 12.0% with a daily increase of up to 19.7%.
The increases in surface O3 were associated with an increase in temperature and PBL height caused by urbanization. Figure 4 shows the changes in annual average surface air temperature and PBL height over the PRD region due to urbanization. Increases in temperature were observed in the PRD region and in urban areas in particular; these increases correlated to increases in photochemical reaction rates and enhanced O3 formation. Our results revealed an increase in both the modeled peroxyacetyl nitrate (PAN) and the ratio of nitrogen dioxide (NO2) to nitric oxide (NO) in urban areas, suggesting that excessive NO2 may react with VOC to form PAN (Hanst, 1971). Moreover, the higher temperature enhanced the temperature-dependent chemical reaction: OH+NO2 → HNO3 (Butkovskaya et al., 2005; Sillman & Samson, 1995). Part of NO2 was hence transformed to HNO3, leading to a higher HNO3/NO2 ratio (Day et al., 2007). This chemical reaction with the PAN chemistry led to less NO to react with O3. On the other hand, despite the reduction in wind spend (Figure S2), the increase in PBL height (100–200 m) caused a dilution of NOx that may weaken NOx titration destruction and thus increased O3 (Li et al., 2016), which was supported by the strong negative correlation (−0.62) between simulated O3 and NOx. Notably, the change in O3 caused by the increase in biogenic VOC emissions in warmer environments was not considered because the emissions were kept constant between the two scenarios to isolate the effect of changes in land cover.
Table 2 demonstrates that changes in O3 varied according to the season and the time of a day (daytime or nighttime). The largest changes in daily average O3 (regional value; urban value) occurred in autumn (1.6%; 30.1%), followed by summer (1.5%; 28.8%), whereas the least changes were observed in winter (1.0%; 11.2%) and spring (0.8%; 14.3%). Moreover, the largest changes occurred primarily at night. In terms of percentage changes, the effect of urbanization on urban O3 was larger in the nighttime than in the daytime; these results were similar to the regional results (Li et al., 2016; Liao et al., 2015). We noted that the ambient ozone concentration was higher in the daytime than in the nighttime and that percentage changes in the nighttime (11.2–30.1%) were thus larger than those in the daytime (6.0–9.7%).
| Daytime | Nighttime | |||
|---|---|---|---|---|
| Regional | Urban | Regional | Urban | |
| Spring | 0.0 | 6.0 | 0.8 | 14.3 |
| Summer | 0.5 | 8.9 | 1.5 | 28.8 |
| Autumn | 0.5 | 9.7 | 1.6 | 30.1 |
| Winter | 0.8 | 9.7 | 1.0 | 11.2 |
3.2 Effect of Urbanization on O3 Sensitivity to NOx and VOC Emissions
3.2.1 Baseline O3 Sensitivity to NOx and VOC Emissions
To understand the effect of urbanization on O3 sensitivity to NOx and VOC emissions, we first analyzed the characteristics of the O3 sensitivities in the heavily urbanized scenario, which were reflective of the air quality in 2010. Figures 5a and 5b display annual average O3 sensitivity to NOx and VOC emissions in 2010. We found that annual average O3 sensitivity to VOC emissions was positive, with a regional average of 0.69 ppb. In major cities, the average O3 sensitivity to VOC emissions was 1.13 ppb, ranging between 0.57 and 1.63 ppb.
Annual average O3 sensitivity to NOx emissions in the PRD region was −3.45 ppb. O3 sensitivity to NOx emissions was typically negative in urban areas but positive in rural areas. The finding of negative sensitivity in urban areas reflected the O3 titration in urban areas (Pun et al., 2003; Yarwood et al., 2003), which experience heavy NOx emissions and thus show a higher NOx-VOC ratio. We estimated O3 sensitivity to NOx emissions for all cities in the PRD region and found the average sensitivity across all cities to be −9.66 ppb. The most negative sensitivity to NOx was found in Foshan (−15.71 ppb), followed by Jiangmen (−13.42 ppb), Guangzhou (−12.82 ppb), and Zhongshan (−12.76 ppb).
Figure 6 depicts the relationship between the NOx/VOC emission ratio and O3 sensitivity to NOx emissions. The regression line indicated that O3 sensitivity to NOx emissions tended to negatively correlate with the NOx/VOC emission ratio. Among the cities in the PRD region, the O3 sensitivities to NOx emission values for Huizhou and Zhaoqing were the least negative, at −3.00 and −4.99 ppb, respectively. This may be the result of their low NOx/VOC emission ratios, which were 3.35 (NOx: 0.072 mole/s; VOC: 0.021 mole/s) and 0.66 (NOx: 0.022 mole/s; VOC: 0.033 mole/s), respectively.
Our results showed seasonal variations in O3 sensitivity to NOx and VOC emissions. On average, the regional O3 was more sensitive to NOx emissions in winter (−5.30 ppb), followed by autumn (−3.68 ppb), spring (−3.47 ppb), and summer (−1.38 ppb). This seasonal pattern differed from that of urban areas, where O3 was more sensitive to NOx in summer (−15.76 ppb) and spring (−14.99 ppb), followed by winter (−12.66 ppb) and autumn (−11.68 ppb). We discerned a spatial pattern for such seasonal variation; for cities closer to the coast (Jiangmen, Zhongshan, Zhuhai, Macau, and Hong Kong), the most sensitive O3 response occurred in autumn, whereas for inland cities (Guangzhou, Dongguan, and Shenzhen), the most sensitive O3 response occurred in summer. This pattern may be associated with the prevailing seasonal wind direction in the region. As shown in Figure S3, the prevailing wind was due northeast in autumn such that the cities closer to the coast (including Jiangmen, Zhongshan, Zhuhai, Macau, and Hong Kong) were in the downwind areas of PRD. These cities therefore had the most sensitive O3 response in autumn. On the other hand, due to the prevailing southerly wind in summer, the inland cities (Guangzhou, Dongguan, and Shenzhen) had the most sensitive O3 response.
3.2.2 Changes in O3 Sensitivity to NOx and VOC Emissions
Figures 5c and 5d depict annual average changes in O3 sensitivity to NOx and VOC emissions caused by urbanization. The mean percentage change in regional O3 sensitivity to NOx in the PRD region was estimated to be −4.0%. Changes in O3 sensitivity to NOx emissions varied at different locations. For example, the percentage change in O3 sensitivity to NOx emissions in Zhongshan was −9.1%, while that in Shenzhen and Foshan was 3.2% and 2.3%, respectively. The average percentage change in regional O3 sensitivity to VOC emissions was negative (−2.4%) with an obvious spatial variation. For example, the percentage change in O3 sensitivity to VOC emissions in Foshan and Shenzhen was 3.8% and 3.1%, respectively, whereas that in Zhongshan was −7.7%. We note that the changes in ozone depend on both changes in climate and background air composition. For example, changes in temperature may change the photolysis rate that would affect ozone. In addition, a unit of NOx emissions adding in a rural area (lower NOx/VOC ratio) or an urban area (higher NOx/VOC ratio) would have different effects. This explains the variations of the impacts over different regions.
Our results revealed a discrepancy in the seasonal variations of changes in regional and urban O3 sensitivity to VOC emissions. Changes in regional O3 sensitivity to VOC emissions were negative in spring (−0.02 ppb), summer (−0.04 ppb), and autumn (−0.01ppb), whereas there was a marginal change in winter. While the regional changes were marginal, changes were profound in urban areas, and changes in urban O3 sensitivity to VOC emissions were all positive, with an average value of 0.56 ppb. The greatest increase in O3 sensitivity to VOC emissions in urban areas occurred in autumn, followed by winter, spring, and summer. We note that larger changes in the urban O3 sensitivity are expected due to the fact that the land cover changes were mainly from nonurban land cover type to urban.
3.3 Effects on Public Health
Table 3 lists the annual average incidences of O3-induced premature mortalities in the PRD region and its major cities. We estimated that 3,964 (95% CI: 2,311–6,172) premature mortalities were caused by ambient O3 exposure in 2010 (heavily urbanized scenario). Zhaoqing (810 [95% CI: 532–1,177]), Jiangmen (616 [95% CI: 386–925]), Guangzhou (614 [95% CI: 293–1,045]), and Huizhou (612 [95% CI: 397–894]) had the highest O3-induced premature mortalities. When combined, the premature mortalities of these four cities accounted for approximately 67.0% of the regional total.
| Heavily urbanized scenario (2010; unit: deaths) | Percentage changes (%) between heavily and lightly urbanized scenarios (absolute changes with unit: deaths) | |
|---|---|---|
| Dongguan | 153 (81–250) | 143.9 |
| Foshan | 346 (175–576) | 105.1 |
| Guangzhou | 614 (293–1,045) | 98.9 |
| Hong Kong | 387 (213–620) | 25.0 |
| Huizhou | 612 (397–894) | 19.0 |
| Jiangmen | 616 (386–925) | 21.1 |
| Shenzhen | 126 (57–218) | 99.8 |
| Zhaoqing | 810 (532–1,177) | 18.1 |
| Zhongshan | 152 (84–243) | 54.9 |
| Zhuhai | 149 (93–224) | 23.2 |
| Regional total | 3,964 (2,311–6,172) | 39.6 |
- Note. The values in parentheses are 2.5 and 97.5 percentiles. The premature mortalities were rounded up to the nearest integer, whereas the percentage changes were rounded to one decimal place, and the absolute changes were rounded to the nearest integer.
Comparison of the two land cover scenarios revealed that the effect of urbanization on O3 caused an increase of approximately 39.6% in O3-induced premature mortalities (1,100 deaths) in the region. Percentage changes in premature mortalities varied by city. The largest percentage changes occurred in Dongguan (143.9%), Foshan (105.1%), Shenzhen (99.8%), Guangzhou (98.9%), and Zhongshan (54.9%). Percentage changes in other cities were ranged between 18.1% and 25.0%.
4 Discussion and Conclusion
Continual population growth and rapid urbanization are expected to cause a persistent increase in the urban population (i.e., the urban population) worldwide. Of that increase, 90% is expected to occur in Africa and Asia; in these two continents, 60% and 52% of the population currently resides in rural areas, respectively (Department of Economic and Social Affairs, Population Division, United Nations, 2006). The projected high growth rate of the urban population is expected to prompt rapid urban expansion.
Studies have reported the urbanization effect on surface ozone in the PRD region (Li et al., 2016; Wang et al., 2007; Wang et al., 2009). Table S9 lists seasonal average changes in O3 reported in both the current study and previous studies. A comparison reveals that our findings are consistent with the findings of other studies and also highlight the effect of urbanization on surface O3. While other studies have focused on a single episode or a month, our study revealed that changes in O3 due to urbanization varied diurnally and by month.
We estimated that exposure to ambient O3 alone caused approximately 4,000 premature mortalities in the PRD region every year. This finding highlights the severe health impact of outdoor air pollution in the region. Furthermore, the subsequent increase in the O3 concentration caused by urbanization led to a 39.6% increase in premature mortality due to O3 exposure per year. Some cities, such as Guangzhou and Foshan, exhibited an increase of approximately 90 and 120 premature mortalities due to urbanization, respectively. From the perspective of public health, the mitigation of O3 pollution in the region is paramount.
To formulate an effective control strategy for O3, better understanding of O3 sensitivity to its precursor emissions is essential. Notably, positive O3 sensitivity to one type of emission indicated that O3 would increase with a unit increase in the corresponding emission type, whereas negative O3 sensitivity indicated an opposite response of O3. Our findings show that baseline regional O3 increased with VOC emissions but decreased with NOx emissions. Moreover, urban O3 was more sensitive to a unit change in NOx emissions than a unit change in VOC emissions, especially in the downwind cities.
Urbanization caused changes in O3 responses to emissions. We noted that changes in O3 sensitivity had different implications because the baseline values of O3 sensitivity may be either positive or negative. When the baseline values of O3 sensitivity are negative, such as the case for urban O3 sensitivity to NOx emissions, the positive absolute change indicates that O3 becomes less sensitive to the corresponding emission changes, whereas negative absolute change indicates the opposite response of O3. For percentage change, the positive values indicate that O3 becomes more sensitive to such changes, whereas the negative value indicates the opposite response of O3. The detail about the implication is provided in Table S8 in SI.
Our sensitivity analysis also revealed the spatial variations in changes in O3 sensitivity to NOx and VOC emissions, implying that O3 concentration may response differently to any emission changes after urbanization. The overall less O3 sensitivity to NOx emissions in urban areas was due to the urbanization-induced increase in temperature that enhanced the temperature-dependent chemical reaction: OH+NO2 → HNO3 (Butkovskaya et al., 2005; Sillman & Samson, 1995). The temperature in urban areas was estimated to be higher in 2010 than in 1990. Part of NO2 was hence transformed to HNO3, leading to a higher HNO3/NO2 ratio (Day et al., 2007). Despite the increase in PAN as a reservoir of NOx, more NOx are needed to result in a same level of O3. For O3 sensitivity to VOC emissions, the overall increase may be due to the fact that in urban areas (VOC-limited), NO2 effectively competes with VOC for OH radicals, leading to a lower production of organic peroxy radicals and hydroperoxyl radicals. With elevated temperature due to urbanization that caused less NO2 through the conversion to HNO3 (Butkovskaya et al., 2005), more OH radicals were available to react with VOC, increasing ozone formation. This explains the increased sensitivity of urban O3 to VOC emissions.
Relevant literature has proposed that climate change mitigation and adaptation strategies may produce potential cobenefits for thermal comfort and air quality (Harlan & Ruddell, 2011). As discussed above, the higher urban O3 sensitivity to VOC emissions after urbanization indicates that urban greenery, which emits VOC (Guenther et al., 2006) and is frequently proposed as a method for mitigating urban climate issues, may deteriorate urban air quality. Unless plant species with low VOC emissions are selected, a trade-off between urban climate and air quality may thus be necessary when considering urban greenery. In addition, the obvious spatial variations in O3 sensitivity to emissions highlight that a same environmental policy may result in different effects. Given the complex and varying interaction of land cover, climate, and air quality, as well as pollutant sensitivity to emissions in each city, we therefore propose a precision environmental management concept that emphasizes the importance of considering the specific atmospheric condition and composition of a city when formulating its environmental policies.
We note that urbanization may increase emissions, especially anthropogenic emissions. This study did not consider increases in emissions due to urbanization because our aim was to investigate the direct effect of urbanization on air quality. Impacts of the increased emissions on air quality and public health should be studied in the future. In addition, previous studies reported that anthropogenic heat is one of the major contributors of temperature changes in urban areas (Chen et al., 2016; Feng et al., 2012; Xie et al., 2016). However, few detailed information or data about anthropogenic heat emissions is available for our model simulations, and the impact of anthropogenic heat therefore still needs further research.
Acknowledgments
This work was jointly funded by The Vice-Chancellor's Discretionary Fund of The Chinese University of Hong Kong (Grant 4930744), the Early Career Scheme of Research Grants Council of Hong Kong (Grant ECS-24301415), the Focused Innovations Scheme of The Chinese University of Hong Kong (project 1907001), and the Environment and Conservation Fund (ECF project 07/2014). We would like to thank the Hong Kong Environmental Protection Department and the Hong Kong Observatory for providing air quality and meteorological data, respectively. We acknowledge the support of the CUHK Central High Performance Computing Cluster, on which computation in this work has been performed. The authors declare no competing financial interest. The meteorological data used in this study are openly available from National Centers for Environmental Prediction/Final (NCEP/FNL) at https://rda.ucar.edu/datasets/ds083.2/. The emission data sets used in this study are available online and openly accessed at http://www.meicmodel.org/. The observations of PRD air quality monitoring sites are openly accessed online from air quality report (https://www.epd.gov.hk/epd/sites/default/files/epd/english/resources_pub/publications/files/PRD_2010_report_en.pdf). The population size data sets are 1 KM Grid GDP Data of China in 2010, which are available at http://www.geodoi.ac.cn/weben/CategoryList.aspx?categoryID=9, and the Gridded Population of the World (GPW) v4, which are available at http://sedac.ciesin.columbia.edu/data/collection/gpw-v4.
