Impact of Urbanization on Meteorology and Air Quality in Chengdu, a Basin City of Southwestern China

Rapid urbanization has the potential to fundamentally perturb energy budget and alter urban air quality. While it is clear that urban meteorological parameters are sensitive to urbanization-induced changes in landscapes, a gap exists in our knowledge about how changes in land use and land cover affect the dynamics of urban air quality. Herein, we simulated a severe O3 episode (10–16 July 2017) and a highly polluted PM2.5 episode (25–30 December 2017) and assessed the changes of meteorological phenomenon and evolution of air pollutants induced by urbanization. We found that the urban expansion area (i.e., land use transition from natural to urban surfaces between 2000 and 2017, UEA) has a significant increase in nocturnal 2-m temperature (T2) with maximum values reaching 3 and 4°C in summer and winter, respectively. In contrast, UEA experienced cooling in the daytime with stronger reductions of T2 in winter than in summer. The T2 variability is primarily attributed to the intense thermal inertia and high heat capacity of the urban canopy and the shadowing effect caused by urbanization. Owing to increased surface roughness and decreased surface albedo as well as shadowing effects, the ventilation index (VI) of UEA increased up to 1,200 m2/s in winter while decreased up to 950 m2/s in summer. Changes in meteorological phenomenon alter physical and chemical processes associated with variations in PM2.5 and O3 concentrations. Urbanization leads to enhanced vertical advection process and weakened aerosol production, subsequently causing PM2.5 levels to decrease by 33.2 μg/m3 during the day and 4.6 μg/m3 at night, respectively. Meanwhile, O3 levels increased by 61.4 μg/m3 at 20:00 due to the reduction of horizontal advection induced by urbanization, while O3 concentrations changed insignificantly at other times. This work provides valuable insights into the effects of urbanization on urban meteorology and air quality over typical megacities, which support informed decision-making for urban heat and air pollution mitigation.


INTRODUCTION
Urbanization is an ubiquitous trend worldwide driven by rapid industrialization and socioeconomic development. More than 60% of the world population resided in urban areas in 2018 and it is projected to increase by 80% by 2030 (Grimm et al., 2008;Seto et al., 2012). Since 2000s, environmental issues including heatwaves, flooding, and poor air quality accompanied by urbanization have become public concerns which puts the environmental quality of cities under pressure (Fang et al., 2022). In particular, air pollution featured by high levels of ozone and particulate matter is recognized as a major environmental issue in highly urbanized megacities and city clusters in Asia. Urbanization leads to dramatic land use and land cover (LULC) alteration, which generally results in lower albedo, higher roughness, and higher thermal inertia (Fan et al., 2017). These changes could significantly impact surface energy budget and further influence regional meteorology (Vahmani and Hogue, 2014). Moreover, the street canyon and the gaps between buildings change the radiative forcing . Changes in radiative forcing can lead to changes in near-surface temperatures, affecting the dispersion of pollutants and having a profound impact on air quality .
Prior studies that examine urbanization effects on local and regional-scale city environment were generally based on coupled urban canopy models (UCM) to mesoscale weather forecast models and air quality models (Lin et al., 2016;Lian et al., 2018). Over Los Angeles, Li et al. (2019) used WRF-Chem model to explore the effects of urbanization on meteorology and air quality and reported that nighttime 2-m temperature (T2) rise by up to 1.7 • C and PM 2 . 5 concentrations fall by up to 2.5 µg/m 3 at night. In South Korea, Kim et al. (2021) found that urbanization in metropolitan Ulsan caused PM 10 concentrations drop by up to 3 µg/m 3 and 5 µg/m 3 in the daytime and nighttime, respectively. Wang K. et al. (2021) found that LULC change in the Beijing-Tianjin-Hebei region had significant effects on meteorological conditions and air quality at urban stations, and UCM can improve the performance of meteorological simulation. With a focus on the Yangtze River Delta (YRD) in eastern China, Liao et al. (2014) investigated the performance of different urban canopy models in YRD by WRF-Chem model and pointed out that urbanization process significantly influences wind speeds (reduced by 0.7-2.6 m/s) across YRD. For Sichuan Basin situated in western China, Wang H. et al. (2021) found that increased roughness had a significant effect on reducing near-surface wind speeds, leading to the accumulation of PM 2 . 5 in urban areas of Chengdu. In prior work, only a few studies have focused on identifying the land cover changes over Chengdu based on satellite image and examined the variations of meteorological parameters over time, while modeling assessment on meteorology and air quality changes attributed to urbanization has been rarely conducted, which limits the understanding of environmental consequences induced by urbanization over time (Yang et al., 2020a).
Chengdu is the capital city of Sichuan province with 16.7 million residents and suffers from severe O 3 pollution in summer and experiences episodes of PM 2 . 5 pollution in winter (Yang et al., 2020b;Wu et al., 2021). Substantial anthropogenic emissions of oxides of nitrogen (NO x ), primary PM 2 . 5 , and volatile organic compounds (VOCs) are emitted from on-road vehicles and industrial infrastructures, significantly contributing to the formation of O 3 and PM 2 . 5 in urban areas (Wu et al., 2022). Based on inventory analysis, the Sichuan Academy of Environmental Sciences (SCAES) estimated that emissions of NO x , primary PM 2 . 5 , and volatile organic compounds (VOCs) over Chengdu in 2017 were 140 × 10 6 , 70 × 10 6 , and 160 × 10 6 kg, respectively (Xu et al., 2020). Therefore, it is of utmost importance to quantify the magnitude changes of meteorological phenomenon and air quality attributed to rapid urbanization in Chengdu, the typical basin city located on western China.
In this study, we explicitly address the impacts of urbanization on meteorology and air quality by integrating urban canopy model within the WRF-CMAQ modeling system over Chengdu city. The model framework includes historical and current urban land cover datasets obtained from Moderate Resolution Imaging Spectroradiometer (MODIS) in combination with advanced urban parameterizations. Comparisons of meteorological parameters and air pollutants levels are made between model simulations configured with different land use and land cover data from MODIS. The findings of this work are useful in improving understanding on how urbanization modulate various aspects of meteorology and subsequently influence urban environment.

WRF-UCM-CMAQ Model Framework and Configurations
The WRFv4.1.1 model (Salamanca et al., 2011;Skamarock et al., 2019) and the CMAQv5.3.1 model (Appel et al., 2021), are both compiled and operated on a server with Linux environment. WRFv4.1.1 model was used to simulate meteorological conditions with meteorological initial and boundary conditions taken from NCEP 1 • × 1 • Final (FNL) reanalysis dataset. As shown in Figure 1, three nested domains were adopted with horizontal resolutions of 27, 9, and 3 km, respectively. The outermost domain covers southwest China and the innermost domain focuses on Chengdu. The physical parameterizations configured to be used in WRF are listed in Table 1. CMAQv5.3.1 was driven by meteorological fields provided by WRF to model trace gases and aerosols. The initial and boundary conditions for CMAQ were derived from the default profiles which represent a clean atmosphere. BVOC emissions were calculated by The Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGANv2.1) (Guenther et al., 2012;Wu et al., 2020). CMAQ was configured with the Carbon Bond chemical mechanism (CB06) and AERO6. The model simulation was performed over a 15 days period from 1 July 2017 to 16 July 2017 (summer period), which covers the evolution of a typical summertime O 3 episode in Chengdu and a 15 days period from 15 January 2017 to 30 January 2017 (winter period), which covers the evolution of a typical wintertime PM 2 . 5 episode in Chengdu. To reduce bias from meteorological drift, the first 10 days in both simulations were treated as spin-up and were not analyzed in this study.

Simulation Scenarios
To investigate the impact of land use and land cover changes induced by urbanization on regional meteorology and air quality over Chengdu, we conducted WRF-UCM-CMAQ simulations with two scenarios: 2000 LULC and 2017 LULC. The LULC data was obtained from the MODIS MCD12Q1 product. As seen in Figure 2, the urban areas of metropolitan Chengdu significantly expanded and there are considerable transitions from suburban areas to urban areas from 2000 to 2017, reflecting the rapid urbanization processes that occurred in Chengdu. The obvious changes between these two scenarios could result in changes in land surface properties and influence BVOC emissions from vegetation. quality control and are further used to assess the model performance of WRF-CMAQ.

Statistical Analysis
To assess the modeling performance, hourly output data from the WRF-CMAQ model was interpolated to the observation site (Figure 3). The model performance was quantified using statistical metrics including Pearson correlation coefficient (R), mean bias (MB), root mean square error (RMSE), normalized mean bias (NMB), the mean fractional error (MFE), and the mean fractional bias (MFB). The criteria values reported in Emery et al. (2017) and Boylan and Russell (2006) are adopted for comparing the statistical metrics. In previous studies, we assessed the model performance on meteorology and found that it can accurately reflect the diurnal pattern and fluctuation. It should be noted that the model performance evaluation on the wintertime meteorology has been described in detail previously , and thus will not be discussed here.
Where M i and O i are the simulated and observed data, respectively.M i andŌ i are the average of the simulated and observed data, respectively. n is the number of samples.

RESULTS AND DISCUSSION
Spatial Patterns Changes of Temperature and Process Drivers Figure 5 presents spatial changes of T2 induced by LULC alteration via urbanization across Chengdu in winter and summer. Evidently, the most prominent effects are seen in metropolitan Chengdu where dominated by the anthropogenic land surface. Suburban and rural areas exhibit slight warming over time with T2 changes less than 1 • C, indicating the impacts of urbanization at regional-scale. In addition, it is found that nocturnal T2 across UEA (i.e., land use transition from natural to urban surfaces between 2000 and 2017) increases significantly with maximum values reaching 3 and 4 • C in summer and winter, respectively. In contrast, UEA experiences cooling in the daytime with a remarkably reduced T2 in winter than in summer. This pattern is mainly caused by intense thermal inertia and high heat capacity of the urban canopy, which is primarily attributed to manmade materials (e.g., pavements and buildings).
Urbanization turns natural land surface into anthropogenic land surface, thus leading to an increase in heat capacity and a reduction in surface albedo. Furthermore, the intense thermal inertia of the urban canopy hinders the cooling process during nighttime. Therefore, the cooling process of the urban surface is slower than that of the natural surface. Meanwhile, the increasing PBLH (Supplementary Figure 2) can also lead to warming because of lower air cooling rates at night (Li et al., 2019).
In the daytime, T2 in UEA decreases by about 2-3 • C in winter due to the higher thermal inertia of the urban canopy while the change in summer is insignificant. This is mainly attributed to the perturbation of energy balance in urban areas induced by buildings within the city, namely, the shadow effects (SE). SE is explained by the low solar zenith angle in the winter period leading to less amount of solar radiation reaching the ground and subsequently resulting in less radiation shaded by the buildings, while the high solar zenith angle in the summer period causes much solar radiation to reach the ground and a large portion of it blocked by buildings. In summer, T2 increases due to the absorption of solar radiation, and the higher thermal inertia of the urban canopy hinders the process of temperature increase during the day. The temperature difference between the urban surface and the natural surface is insignificant because of the minor SE due to the high solar zenith angle in the summer period. In winter, while the urban canopy has thermal inertia, SE is more pronounced and shade most of the radiation, offsetting the warming and causing the temperature to decrease by 2-3 • C over the UEA.

Spatial Patterns Changes of Ventilation and Process Drivers
Changes in ventilation conditions due to urbanization are discussed in this section. The changes in PBLH affect the vertical diffusion capacity of air pollutants, while the changes in WS10 affect the horizontal transport of ground-level air pollutants. Here, we use the ventilation index (VI) calculated as the product of PBLH and WS10 for illustrating the variations of diffusion capacity over Chengdu (Li et al., 2019). The equation for VI can be written as follow.

Ventilation index = PBLH × WS10
where PBLH is the planetary boundary layer height (m) and WS10 is the 10-m wind speed (m/s). Figure 6 presents spatial changes of VI induced urbanization across Chengdu in winter and summer. In winter, it is found that VI across UEA increases significantly with maximum values reaching 1,400 m 2 /s and 1,300 m 2 /s during the daytime and nighttime, respectively. In contrast, UEA has a significant diurnal variation of VI in summer, with a significant decrease in VI during the nighttime (maximum of 1,400 m 2 /s) and a significant increase during the daytime (maximum of 1,500 m 2 /s).
In winter, during the 00:00-06:00 period ( Supplementary  Figure 4), UEA exhibited stronger WS10 (Supplementary Figure 3) and higher PBLH (Supplementary Figure 2) in 2017, which can be attributed to the mechanism that SE causes longwave radiation in the street valley to be repeatedly reflected and absorbed at night. The urban surface temperature increased due to the accumulation of long-wave radiation and anthropogenic heat rejection, which increases the aerodynamic energy of atmospheric turbulence and increase the WS10 and PBLH. Increasing PBLH is also related to increased surface roughness since shear production dominates turbulent kinetic energy (TKE) at night (Li et al., 2019). During 12:00-18:00 ( Supplementary  Figure 4), WS10 increased by 0.24 m/s and PBLH slightly increased by 47.3 m. The decrease of PBLH is mainly associated with air temperature changes because buoyancy production dominates TKE during the day. Reduced buoyancy production of TKE occurs when air temperature decreases (Figures 5C,D), resulting in a shallower PBLH. Ultimately, the increased WS10 resulted in a 1,400 m 2 /s increase at 12:00 and a 1,300 m 2 /s increase at 18:00 in VI.
In summer, during the 00:00-06:00 period (Supplementary Figure 5), UEA shows weaker WS10 (Supplementary Figure 3) and higher PBLH (Supplementary Figure 2) in 2017. The decrease in WS10 is mostly caused by the dragging force, which increases as the friction effect enhances. Additionally, SE causes the urban surface to heat up due to multiple reflections and absorptions of long-wave radiation in the street valley, delaying the heat dissipation process, and subsequently leading to an increase in PBLH. Consequently, the reduction of WS10 caused the VI to decrease by 500 m 2 /s at 00:00 and 1,200 m 2 /s at 06:00 (Supplementary Figure 5), which is beneficial to confine pollutants near the surface. At 12:00 (Supplementary Figure 5), the increased air temperature enhances the atmospheric turbulence and drives the daytime PBL  growth, resulting in an increase of 1,300 m 2 /s in VI. In summary, WS10 acts as the dominant factor in altering VI in summer, while variability in PBLH governs the changes of VI in daytime and WS10 contributes to variations of VI at night in winter. Figure 7 shows spatial distribution of PM 2 . 5 concentrations in 2000 and 2017 LULC at 00:00, 06:00, 12:00, and 18:00. There are PM 2 . 5 hotspots over UEA and southern Chengdu at 00:00, with peak levels spike to 313.4 and 283.3 µg/m 3 in 2017 and 2000, respectively. In the morning (06:00), PM 2 . 5 levels over downtown Chengdu drop to 126.3 µg/m 3 and elevated PM 2 . 5 concentrations are found over rural areas. It is worth noting that urbanization leads to considerable decrease of PM 2 . 5 concentrations (>40 µg/m 3 ) across UEA at night. However, the change of PM 2 . 5 is insignificant in the daytime over the study domain, implying varied effects induced by urbanization for different time periods.

Temporal and Spatial Patterns of PM 2 . 5 Concentration Changes and Process Drivers
To further examine the dominant contributor that leads to variations of PM 2 . 5 , the chemical components of PM 2 . 5 are classified into the following three categories: secondary inorganic aerosols (SIA), including nitrate (NO 3 − ), sulfate (SO 4 2− ), and ammonium (NH 4 + ); carbonaceous aerosols, including elemental carbon (EC) and primary organic carbon (POC); and secondary organic aerosol (SOA), including SOA formed from anthropogenic VOC precursors (ASOA) and biogenic VOC precursors (BSOA). Figure 8 depicts the diurnal changes of PM 2 . 5 components and total PM 2 . 5 concentrations over UEA. The primary carbonaceous aerosols have similar temporal changes with WS10. During the daytime, the slight changes of WS10 do not substantially affect primary carbonaceous aerosols concentrations. In addition, the shift of land cover dominated by vegetation to anthropogenic land surface leads to reductions of BVOC emissions and subsequently results in reduced BSOA concentrations (by 3.4 µg/m 3 ). Some semivolatile compounds species in ASOA increase by 2.85 µg/m 3 due to the changes in gas-particle phase partitioning from the temperature reduction. The concentrations of nocturnal nitrate and sulfate aerosols decreased by 2.8 µg/m 3 and 1.3 µg/m 3 , respectively. It is worth mentioning that reductions of nitrate and sulfate account for approximately 40% of the decline in PM 2 . 5 concentrations. As sulfate is nonvolatile, weakened atmospheric oxidation rates induced by reduced temperature and enhanced ventilation are dominant contributors to declined sulfate. Unlike large reductions in nitrate and sulfate, CMAQ predicts the changes of NH 4 + to a lesser extent due to relatively low level of ammonia emissions in Chengdu.
The impact of urbanization on PM 2 . 5 concentrations in winter results exhibited the PM 2 . 5 concentrations over UEA significantly increased at night while slightly decreased during the daytime. In 2017, AERO process had a daytime impact of −28.3 µg/m 3 , compared to −19.4 µg/m 3 in 2000 (Figure 9). AERO process enhanced since the temperature decreases during the day, reducing air oxidation rate and hence the SOA level decreased. CHEM and DDEP processes were insignificantly affected by urbanization, while HADV and ZADV processes had remarkable impacts from urbanization due to the increase of VI. HADV process in 2017 resulted in PM 2 . 5 concentrations increased by 26.3 µg/m 3 at night and increased by 5.2 µg/m 3 during the day, whereas HADV process in 2000 resulted in PM 2 . 5 concentrations increased by 13.8 µg/m 3 at night and decrease by 2.2 µg/m 3 during the day. According to ZADV process, PM 2 . 5 concentrations over UEA decreased by 33.4 µg/m 3 and 29.3 µg/m 3 during the nighttime in 2017 and 2000, respectively. This discrepancy could be attributed to the increased T2 from more anthropogenic heat and the higher heat storage capacity, which enhanced vertical convection. Because of the lower WS10 during the nighttime in 2017, HDIF process led to a 0.15 µg/m 3 increase in PM 2 . 5 concentrations, whereas HDIF caused PM 2 . 5 concentrations to decrease by 0.15 µg/m 3 during the night in 2000 due to the higher WS10. In addition, owing to the increase in T2, the UEA acquired a higher temperature gradient, making VDIF process stronger in 2017 than it was in 2000. Figure 10 shows the spatial distribution of simulated O 3 at 00:00, 06:00, 12:00, and 18:00 during the summer episode. It is worth  noting that the Mountain Range area situated in the northwestern Chengdu exhibited elevated O 3 levels between 00:00 and 06:00 with peak O 3 concentrations spiking above 200 µg/m 3 , which is mainly attributed to weak NO x titration at night due to the relatively low anthropogenic NO x emissions over this region. In contrast, O 3 levels in metropolitan Chengdu showed a typical diurnal pattern which is characterized by peak concentrations occurring in the afternoon and decreasing to a low level (less than 30 µg/m 3 ) at night due to sufficient NO x titration.

Temporal and Spatial Patterns of O 3 Concentration Changes and Process Drivers
In comparison to the difference in O 3 concentrations between 2000 and 2017, O 3 levels in UEA increase significantly at midnight (00:00) with an enhancement magnitude reaching  40 µg/m 3 , which is consistent with the variation of nocturnal temperature in Supplementary Figure 5B. This pattern implies that the impacts of the thermal inertia of urban buildings govern this phenomenon, as the cooling rates over anthropogenic land surface are lower than natural LULC. Furthermore, there is a notable increase in O 3 concentrations at 18:00 throughout the UEA with a similar spatial pattern and magnitude as 00:00. Figure 11 demonstrates that O 3 and NO 2 concentrations increased significantly at 20:00 with values of 61.4 and 62.3 µg/m 3 , respectively, while the concentrations of O 3 and NO 2 did not change significantly during the daytime, which is due to that VI had remarkable increase around 20:00 while VI changed marginally during the daytime. Thus, regions with lower NO 2 levels (i.e., UEA) would accumulate more O 3 concentrations due to the weaker titration during the nighttime.
Changes in O 3 concentrations are mainly caused by HADV and VDIF processes, as seen in Figure 12. Accumulated contribution from VDIF increase to 323.4 µg/m 3 becasuse PBLH and vertical wind speed both increase induced by urbanization. In terms of HADV process, the increase of WS10 facilitates the dispersion and transport outflow of O 3 , which cause O 3 concentrations to decrease by 133.7 µg/m 3 in 2017 while to increase 43.3 µg/m 3 in 2000. Higher plant fraction in 2000 leads to stronger DDEP process during the daytime because vegetation could consume O 3 through plant stomata. However, during the nighttime, the intensity of DDEP process in 2000 is weaker than that in 2017 which is due to that O 3 concentrations in 2000 is much lower than that in 2017 at this time.

CONCLUSION
In this study, we use parallel WRF-UCM-CMAQ modeling with LULC data in 2000 and 2017 to investigate the effects of urbanization on regional meteorology and air quality in Chengdu, a typical megacity in the southwestern China. Changes in key meteorological parameters and air pollutants levels are compared in a summertime high O 3 event and a winter PM 2 . 5 episode.
Our results find that land surface alteration induced by urbanization exhibits profound effects on meteorological parameters across UEA. In winter, urbanization leads to reduced T2 (by −1.1 • C) and increased VI (by 819 m 2 /s) during the daytime, while increasing T2 (by 0.46 • C) and VI (by 328 m 2 /s) at night in summer, urbanization increased T2 throughout the day (0.55 • C during the day and 2.0 • C at night) while VI increased by 1,161 m 2 /s during 18:00-20:00 and decreased by 355 m 2 /s during the other times. The magnitude difference in nocturnal T2 between winter and summer is mainly because the intense solar radiation in summer lead urban surface to store more heat, which can raise T2 at night. The differences in VI can be attributed to changes in PBLH and WS10 caused by increased surface roughness and reduced surface albedo, as well as energy changes in atmospheric turbulence caused by SE. Changes in regional meteorology have an impact on gaseous pollutants concentrations and particulate pollution. PM 2 . 5 concentrations are dictated by the various competing processes, including (a) the gas-particle phase partitioning, (b) the atmospheric oxidation rate, and (c) ventilation. In addition, urbanization caused the LULC from vegetation dominated shift to the anthropogenic land surface, which results in reduced BSOA concentrations (by 3.4 µg/m 3 ). These processes lead to PM 2 . 5 decrease by 33.2 and 4.6 µg/m 3 during the daytime and nighttime, respectively. For O 3 , it is found that urbanization leads to insignificant changes in daytime O 3 levels while nocturnal O 3 concentrations in 2017 substantially increase as compared with 2000, which is attributed to VDIF process at night bringing in more concentrations.
Here, we demonstrate the crucial effects of urbanization on urban meteorological parameters and its subsequent influence on air quality by taking the megacity of Chengdu as an example. The findings of this work contribute insight into developing climate adaptation and air pollution mitigation strategies for megacities and heavily urbanized areas.

AUTHOR CONTRIBUTIONS
HW: conceptualization, methodology, formal analysis, investigation, writing-original draft, visualization, and project administration. ZL: conceptualization and project administration. KW: conceptualization and writing-review and editing. JQ: review and editing. YZ: conceptualization, writingreview and editing, and project administration. BY and MH: resources and data curation. All authors contributed to the article and approved the submitted version.