Application of a Modeling Framework to Mitigate Ozone Pollution in Changzhou, Yangtze River Delta Region

Abstract

Ozone pollution in densely populated urban regions poses a great threat to public health, due to the intensive anthropogenic emissions of ozone precursors and is further aggravated by global warming and the urban heat island phenomenon. Air quality models have been utilized to formulate and evaluate air pollution control strategies. This study presents a comprehensive modeling assessment of ozone mitigation strategies during an ozone pollution episode in Changzhou, an industrial city in the Yangtze River Delta region. Utilizing the Community Multiscale Air Quality Modeling System (CMAQ), we quantified the contribution of ozone from different emission sectors and counties within Changzhou using the integrated source apportionment method (ISAM). During the pollution period, local emissions within Changzhou account for an average of 41.5% of MDA8 ozone, with particularly notable contributions from Jingkai (11.2%), Wujin (9.5%), and Liyang (7.8%). Upon these findings, we evaluated three sets of emission reduction scenarios: uniform, sector-specific, and county-specific reductions. Results show that industry and transportation are responsible for over 20% of ozone concentrations, and targeted reductions in these sources yielded the most significant decreases in ozone levels. Notably, reducing industrial emissions alone decreased ozone concentrations by 3.2 μg m⁻³ during the pollution episode. County-specific reductions revealed the importance of targeted strategies, with certain counties showing more pronounced responses to emission controls. On a daily basis, emission reductions in Xinbei contributed to a maximum ozone decrease of 4.4 μg m⁻³. This study provides valuable insights into the efficacy of different mitigation measures in Changzhou and offers a practical and useful framework for policymakers to implement strategies while addressing the complexities of urban air quality management.

1. Introduction

Ground-level ozone (O₃) pollution in densely populated urban areas presents a great threat to public health [1,2,3,4], particularly affecting the growing elderly population that is more vulnerable to the adverse impacts of ozone exposure [5,6]. Driven by global warming, rising atmospheric temperatures typically enhance ozone formation [7,8]. Moreover, the urban heat island effect introduced by urbanization can indirectly aggravate ozone pollution in urban areas by altering local meteorological conditions [9,10,11]. These climate–urbanization feedback loops create complex sustainability challenges requiring integrated mitigation approaches. Over the past decade, the fine particulate matter (PM_2.5) concentrations in China have notably declined at an average rate of 1.45 μg·m⁻³ per year from 2013 to 2022 [12]. In contrast, O₃ concentrations have exhibited considerable fluctuations [13,14,15]. According to the China Blue Book of Atmospheric Ozone Pollution Prevention and Control [16], ozone levels in China from 2015 to 2022 initially exhibited a trend in fluctuating increase, followed by a phase of sustained high variability.

Three-dimensional air quality models, which simulate the physical and chemical processes of air pollution, have been extensively used in air quality management and planning [17,18,19,20]. For example, these models provide a technical foundation for the development of State Implementation Plans (SIPs) by offering data-driven insights into air quality trends, pollutant behavior, and the effectiveness of control measures [21,22]. Cheng et al. [17] integrate emissions, atmospheric, and health models to discuss how a synergistic strategy for air pollution control and carbon neutrality in China could prevent millions of premature deaths annually by 2060. In addition to formulating long-term control policies, air quality models also facilitate the evaluation of short-term control strategies aimed at mitigating peak pollution during heavy pollution periods [23,24]. For instance, Wang et al. [23] proposed a framework for the Joint Regional Prevention and Control (RJPC) program to eliminate a heavy PM_2.5 pollution episode within the Fenhe Plain.

Numerous studies have focused on exploring effective strategies to alleviate ozone pollution across different regions in China. Zhang et al. [25] integrated ground-based measurements with model simulations to analyze changes in ozone formation sensitivity during a persistent pollution episode in Northeast China and explore its implications for developing effective ozone control strategies. Wang et al. [26] evaluated the ozone reduction potential in the Pearl River Delta region through O₃-NOx-VOC sensitivity diagnosis and source apportionment. Results indicated that reducing precursor emissions in Dongguan contributes to improved air quality in the downwind area of Shenzhen. Tian et al. [27] quantified the contributions of cross-emission sources (between urban areas and sectors) to ozone concentrations and implemented targeted controls on anthropogenic emissions in the Central Plains urban agglomeration. Li et al. [28] simulated the response of ozone pollution in Lanzhou to precursor emissions and designed 11 emission control scenarios. Results show that a 10% reduction in NOx combined with a 20% reduction in total VOCs would lead to a decrease in ozone concentrations by up to 13% in urban Lanzhou and 16% in the western suburbs. However, despite these advances, there remains a critical gap: the absence of a unified framework that integrates source apportionment, scenario analysis, and policy evaluation to deliver actionable guidance for local policymakers.

This study addresses this gap by developing and applying a comprehensive modeling framework to evaluate ozone mitigation strategies. Using Changzhou—a typical industrial city in the Yangtze River Delta (YRD) region—as a case study, we demonstrate the framework’s practical utility and effectiveness. Changzhou, with its 5.4 million permanent residents and 1.7 million registered vehicles [29], serves as an important modern manufacturing hub characterized by a robust industrial base and diversified industries [30]. Developing effective ozone control strategies here is crucial for sustaining economic activity while safeguarding population health. In recent years, maximum daily 8 h average (MDA8) O₃ concentrations in Changzhou have followed an upward trend, with the highest ozone levels and exceedance days (MDA8 O₃ > 160 μg·m⁻³) occurring during summer months (June, July, and August). For instance, in June 2022, ozone concentrations exceeded the national standard (MDA8 O₃ = 160 μg·m⁻³) on 43% of days, peaking at 248.9 μg·m⁻³, posing a severe public health risk. Previous studies have explored the mechanisms of ozone formation in Changzhou [31,32,33], revealing insights such as the shift in ozone generation mechanisms during the 2020 COVID-19 lockdown [32] and the influence of local chemical reactions and Lake Tai’s lake–land breeze on near-surface O₃ levels [31]. Nevertheless, the previous studies have not provided a systematic analysis of ozone contributions from different sectors and regions within Changzhou, and there has been a lack of comprehensive evaluation of various mitigation strategies tailored to the city of Changzhou.

This study employs the widely recognized Community Multiscale Air Quality Modeling System (CMAQ) to present a comprehensive modeling framework for evaluating mitigation strategies during a severe ozone pollution episode in Changzhou. Using CMAQ’s integrated source apportionment method (ISAM), we quantify ozone contributions from different emission sectors and counties within Changzhou. Based on these findings, we develop multiple emission reduction scenarios, including uniform, sector-specific, and county-specific reductions, and evaluate their effectiveness in lowering ozone concentrations. Our sector- and county-specific reduction scenarios identify pathways to maximize ozone reduction, supporting resource-efficient sustainability planning. Our study provides valuable insights into the efficacy of different ozone mitigation measures during ozone pollution episodes in Changzhou and also offers a practical and useful framework for policymakers to implement strategies that protect public health while addressing the complexities of urban air quality management.

2. Materials and Methods

2.1. Methodological Framework

As shown in Figure 1, a framework is proposed for designing the optimal emission reduction strategies to mitigate O₃ pollution in the target city. The process starts with the simulation of a base case scenario utilizing meteorology and air quality models. Observations of meteorological variables and O₃ are employed to assess model performance. Subsequently, ozone source apportionment techniques are applied to quantify ozone contributions from diverse aspects, such as regional transport and surrounding cities, inter-county contributions within the target city, and emissions from various local source categories. Based on these source apportionment outcomes, a range of emission reduction scenarios are explored using the brute-force method. These scenarios encompass different NOx and VOC reduction ratios, county-specific emission reductions, and sector-specific emission reductions. The effectiveness of these scenarios in reducing O₃ concentrations is evaluated to pinpoint a tailored ozone mitigation strategy for the target city.

2.2. Modeling Configuration

In this study, the Weather Research and Forecasting (WRF) model (version 4.0, https://www.mmm.ucar.edu/models, accessed on 11 November 2024) in conjunction with CMAQ version 5.3.2 [34] was used. The WRF and CMAQ configurations employed in this study were consistent with our previous studies [35], which are shown in

Table A8. Parameter settings for the WRF and CMAQ models.

Simulation Region
Horizontal resolution	36 km, 12 km, and 4 km
Center longitude and latitude	34° N, 110° E
Projection method	Lambert conformal conic
WRF
Microphysical process	Purdue–Lin
Land surface process	Noah land surface model
Planetary boundary layer scheme	Yonsei University
Long-wave radiation	RRTM
Short-wave radiation	Goddard
Cumulus convection scheme	Kain–Fritsch
CMAQ
Gas-phase chemical mechanism	CB6
Aerosol module	AERO7
Initial/boundary conditions	CMAQ default profile

. The modeling domain consists of three nested Lambert projection grids (Figure 2a) with spatial resolutions of 36 km, 12 km, and 4 km. Changzhou is located within the innermost grid, which encompasses the entire YRD region, including Anhui, Jiangsu, Zhejiang, and Shanghai. Boundary conditions for D01 were obtained from the default CMAQ profile, whereas those for D02 and D03 were derived from the outputs of their corresponding upper-level simulation domains. We employed the Carbon Bond mechanism (CB06) along with the AERO7 module to simulate gas-phase chemistry and aerosol formation. For anthropogenic emissions, we utilized the Multi-Resolution Emission Inventory of China (MEIC) developed by Tsinghua University (http://meicmodel.org, accessed on 11 November 2024), except in the YRD region, where anthropogenic emissions were consistent with those used in our prior studies [36,37]. The YRD emission inventory included emissions from power plants, industrial boilers and kilns, industrial processes, transportation, solvent use, oil and gas storage, residential, agriculture, dust, and biomass burning. Annual total emissions were allocated into hourly emissions with a spatial resolution of 4 km using the Sparse Matrix Operator Kernel Emissions (SMOKE) processing system [38]. Biogenic VOC emissions were calculated using the Model of Emissions of Gases and Aerosols from Nature (MEGAN version 3.2; https://bai.ess.uci.edu/megan, accessed on 1 November 2024). The specific ozone pollution episode selected for simulation spanned from 14 to 21 June 2022. During this 8-day period, the average MDA8 O₃ concentrations reached 202.3 μg·m⁻³ with a peak concentration of 248.9 μg·m⁻³ observed on 18 June. The winds in Changzhou shifted from easterly to northerly during the first four days and then transitioned to persistent easterly winds starting from 18 June (Figure A17). The model was initialized with a five-day spin-up period to reduce the influence of initial conditions.

The ISAM module in the CMAQ model was applied to quantify the sectoral and county ozone contributions in Changzhou. For sectoral contributions, we identified six categories of emission sources: transportation, industrial processing, industrial boilers and kilns, power plants, natural emissions, and other sources (including biomass burning, residential emissions, solvent usage, oil and gas storage, dust, and shipping activities). In terms of county-level contributions, we quantified ozone contribution from each of the seven counties within Changzhou (Figure 2c): Jintan (JT), Xinbei (XB), Tianning (TN), Wujin (WJ), Jingkai (JK), Liyang (LY), and Zhonglou (ZL). For each county, the simulated ozone concentrations at the national monitoring station (if not available, the provincial monitoring station was used) were extracted and averaged to represent the county-level ozone concentrations.

2.3. Ozone Mitigation Scenarios

Although the ISAM results provided valuable information on ozone source contribution in Changzhou, the tagging-based method is not suitable for supporting air quality planning of non-linear species, like O₃ [39,40]. To support air quality planning, three sets of ozone mitigation scenarios were developed to investigate their efficacy in reducing ozone levels using the brute-force method. These scenarios were categorized as “uniform emission reduction”, “sector-specific reduction”, and “county-specific reduction” (

Table 1. Summary of different mitigation scenarios (see Table A2, Table A3 and Table A4 for details).

Mitigation Scenario	NOx/AVOC Emissions	Sectors	Counties	Simulations
Uniform emission reduction	Reduction ratio (20%, 40%, 60%) applied to NOx and/or AVOC	All sectors	All counties	Case 1–Case 9
Sector-specific reduction	Only reducing AVOC by 40%	One sector at a time	All counties	Case 10–Case 14
County-specific reduction	Only reducing AVOC by 40%	All sectors	One county at a time	Case 15–Case 21

). In the “uniform emission reduction” scenarios, a consistent reduction ratio was applied across all emission sectors and counties within Changzhou. Various combinations of reduction ratios (20%, 40%, 60%) were tested for anthropogenic AVOC (AVOC) and anthropogenic NOx (ANOx), both separately and simultaneously, leading to a total of nine simulations conducted. In the “sector-specific reduction” scenarios, the emission reduction ratio was applied to one sector at a time while keeping other sectors unchanged (a total of five simulations). Similarly, in the “county-specific reduction” scenarios, the emission reduction ratio was applied to one county at a time, with emissions from other counties remaining unchanged (a total of seven simulations). Building upon the results from the “uniform emission reduction” scenarios, a 40% reduction ratio was exclusively applied to AVOC in both the sector-specific and county-specific scenarios. For emissions outside of Changzhou, we assumed a 20% emission reduction in NOx and AVOC to reflect the emission reduction efforts at the regional level.

2.4. Model Performance Evaluation

The model performance of baseline simulation (i.e., with no emission reductions) was evaluated by comparing simulated meteorological variables and ozone concentrations to observations. Several statistical metrics were used for model performance evaluation, including mean bias (MB), normalized mean bias (NMB), correlation coefficient (R), and index of agreement (IOA) (see

Table A5. Definition of model performance evaluation metrics used in this study.

No.	Statistics Index	Definition
1	Mean bias (MB)	$MB={\displaystyle \frac{1}{N}}\sum \left(M_i-O_i\right)$
2	Normalized mean bias (NMB)	$NMB=100\%\times {\displaystyle \frac{\sum \left(M_i-O_i\right)}{\sum O_i}}$
3	Correlation coefficient (R)	$R={\displaystyle \frac{{\sum }_1^N\left(\left(M_i-\overline{M}\right)\times \left(O_i-\overline{O}\right)\right)}{\sqrt{{\sum }_1^N{\left(M_i-\overline{M}\right)}^2{\sum }_1^N{\left(O_i-\overline{O}\right)}^2}}}$
4	Index of agreement (IOA)	$IOA=1-{\displaystyle \frac{{\sum }_1^N{\left(M_i-O_i\right)}^2}{{\sum }_1^N{\left(\left|M_i-\overline{O}\right|+\left|O_i-\overline{O}\right|\right)}^2}}$

for detailed formulae). Hourly observed O₃ concentration was obtained from the Ecological Environment Data Platform (https://quotsoft.net/air, accessed on 1 November 2024). Meteorological variables, including 2 m temperature (T), relative humidity (RH), and 10 m wind speed (WS) and wind direction, were from the National Meteorological Information Center (http://data.cma.cn/, accessed on 1 November 2024).

Simulated meteorological parameters were compared with observations in the YRD region (

Table A6. Model performance evaluation of WRF-simulated 2 m temperature (T₂), 10 m wind speed (WS₁₀), and relative humidity (RH).

City	Meteorological Parameters	Observation	Simulation	MB	NMB	R	IOA
Changzhou	T2 (°C)	28.0	27.8	−0.1	−0.4%	0.8	0.9
	WS10 (m/s)	3.4	3.7	0.4	10.7%	0.4	0.6
	RH (%)	66.3	64.0	−2.4	−3.6%	0.8	0.9
Hangzhou	T2 (°C)	26.9	27.7	0.8	3.0%	0.8	0.9
	WS10 (m/s)	2.8	3.1	0.3	12.0%	0.4	0.6
	RH (%)	77.9	66.7	−11.1	−14.0%	0.8	0.8
Shanghai	T2 (°C)	26.6	26.1	−0.5	-2.0%	0.9	0.9
	WS10 (m/s)	3.7	3.8	0.1	2.0%	0.4	0.7
	RH (%)	78.6	72.3	−6.2	−8.0%	0.8	0.9
Nanjing	T2 (°C)	27.7	28.7	1.0	4.0%	0.8	0.9
	WS10 (m/s)	3.8	3.9	0.1	2.0%	0.5	0.7
	RH (%)	66.4	59.6	−6.8	−10.0%	0.8	0.8
Hefei	T2 (°C)	28.0	28.4	0.4	1.0%	0.8	0.9
	WS10 (m/s)	3.6	4.0	0.4	11.0%	0.4	0.6
	RH (%)	70.5	64.7	−5.8	−8%	0.8	0.8

). Overall, the WRF model effectively captured the observed variations in T, WS, and RH in Changzhou. WS was slightly overestimated with an MB of 0.4 m/s and RH was underestimated with an NMB of −3.6%. Based on the results of the four provincial capital cities (Shanghai, Hangzhou, Nanjing, and Hefei), the simulated T and RH showed a strong correlation with the observed data, with R exceeding 0.8. WS was generally overestimated, with an MB ranging from 0.1 to ~0.4 m/s. However, RH tended to be underestimated with an NMB ranging from −14% to −8%. The overestimated wind speeds could lead to underestimated O₃ concentrations due to the dispersion of O₃ precursors and enhanced horizontal and vertical mixing of the atmosphere [41]. On the other hand, underestimated humidity could somehow offset the O₃ underestimation as a negative correlation was usually found between humidity and O₃ [42,43].

The validation of simulated O₃ concentration is shown in Figure 3 and

Table A7. Model performance evaluation for simulated MDA8 O₃ in 41 cities in the YRD region during June 2022.

City Name	Observation (µg·m−3)	Simulation (µg·m−3)	MB (µg·m−3)	NMB (%)	RMSE	R	IOA
Hangzhou	129.5	117.2	−12.3	−9.5	39.0	0.7	0.8
Huzhou	139.6	129.3	−10.3	−7.4	39.5	0.6	0.7
Jiaxing	137.7	136.9	−0.8	−0.6	39.4	0.5	0.7
Jinhua	108.9	101.3	−7.6	−7	34.1	0.8	0.8
Lishui	101.6	71.0	−30.6	−30.1	38.4	0.7	0.7
Ningbo	102.4	99.3	−3.1	−3	38.7	0.5	0.7
Quzhou	100.0	89.4	−10.6	−10.6	27.3	0.8	0.8
Shaoxing	123.0	111.3	−1.7	−9.5	35.1	0.7	0.8
Taizhou (Zhejiang)	96.2	91.2	−5.0	−5.2	30.1	0.7	0.8
Wenzhou	99.2	89.6	−9.6	−9.3	33.2	0.7	0.8
Zhoushan	73.4	80.3	7.0	9.5	28.0	0.5	0.7
Shanghai	121.5	121.8	0.3	0.3	37.5	0.6	0.7
Changzhou	134.8	148.3	13.6	10.1	46.5	0.5	0.6
Huai’an	128.2	150.3	22.1	17.3	47.8	0.4	0.6
Lianyungang	122.4	144.7	22.3	18.2	38.9	0.7	0.8
Nanjing	143.2	141.2	−2.0	−1.4	43.0	0.5	0.7
Nantong	117.9	141.8	24.0	20.3	47.8	0.6	0.8
Suqian	142.0	159.9	17.9	12.6	41.7	0.5	0.7
Suzhou (Jiangsu)	137.7	136.6	−1.1	−0.8	38.9	0.6	0.8
Taizhou (Jiangsu)	131.5	135.7	4.2	3.2	46.6	0.4	0.6
Wuxi	134.8	145.7	10.9	8.1	40.6	0.6	0.8
Xuzhou	147.4	168.4	21.0	14.2	50	0.3	0.5
Yancheng	121.1	148.1	27.0	22.3	49.6	0.6	0.7
Yangzhou	140.9	151.5	10.5	7.5	52.7	0.4	0.6
Zhenjiang	138.5	151.9	13.4	9.7	49.4	0.5	0.7
Anqing	115.0	105.2	−9.8	−8.5	37.5	0.6	0.7
Bozhou	145.4	164.1	18.7	12.8	50.1	0.1	0.4
Chizhou	118.5	104.0	−13.5	−11.4	35.4	0.7	0.8
Chuzhou	145.9	148.7	2.8	1.9	40.9	0.6	0.7
Fuyang	142.8	146.1	3.3	2.3	47.0	0.2	0.5
Hefei	132.5	127.0	−5.5	−4.1	37.0	0.3	0.6
Huaibei	147.0	157.1	10.1	6.9	45.0	0.2	0.5
Huangshan	102.5	82.8	−19.7	−19.2	29.2	0.8	0.7
Lu’an	128.6	122.7	−5.8	−4.5	41.2	0.3	0.6
Ma’anshan	140.1	139.1	−1.0	−0.7	37.6	0.5	0.7
Suzhou (Anhui)	141.2	152.6	11.4	8.1	42.9	0.3	0.6
Tongling	109.0	105.3	−3.7	−3.4	34.4	0.6	0.8
Wuhu	124.6	121.1	−3.5	−2.8	32.9	0.6	0.8
Xuancheng	118.6	92.9	−25.6	−21.6	36.7	0.7	0.7
Bengbu	146.9	154.0	7.2	4.9	42.1	0.3	0.6
Huainan	140.6	139.7	−1.0	−0.7	40.5	0.2	0.6

. In general, the CMAQ model was able to capture both the spatial and temporal variations in observed O₃. During the selected pollution period (Figure 3b), the model showed an NMB of −1.0% and a spatial correlation coefficient of 0.9. Ozone underestimation was observed in northern and central Jiangsu, with MB ranging from 4.2 to ~27.0 µg·m⁻³. Furthermore, we compared the hourly O₃ simulated concentrations for Changzhou and four provincial capitals in the YRD region (Figure 3c). During June 2022, R values ranged from 0.6 to 0.7 for these five cities. During the pollution period, the observed MDA8 O₃ concentration in YRD region was 149.5 µg·m⁻³, whereas the simulated concentration was 147.5 µg·m⁻³. In summary, these results indicated acceptable model performance of the WRF and CMAQ model for further assessment of the efficacy of different emission reduction scenarios.

3. Results

3.1. Ozone Source Apportionment Based on CMAQ-ISAM

The regional and sectoral contributions of MDA8 during the selected pollution episode are shown in Figure 4. Local emission sources within Changzhou accounted for an average of 41.5% of MDA8 ozone, which was slightly higher than the contribution from boundary conditions (36.9%). On a daily basis, local sources could contribute as much as 75% (on 21 June). On a county-level (Figure 4a), the three counties with the highest contributions during the pollution episode were JK (11.2%), WJ (9.5%), and LY (7.8%), collectively accounting for ~28.5% of the total ozone contribution. The remaining counties (ZL, XB, TN, and JT) each contributed less than 5%, underscoring the spatial heterogeneity in emission impacts.

The average contributions from neighboring cities during the ozone episode amounted to 15.7%, and the relative importance of these surrounding cities varied primarily in response to changes in wind direction, with Wuxi (7.3%) exhibiting the highest contributions to ozone levels in Changzhou on average (Figure 4c,d). Compared to the monthly averaged results (Figure A2), the contributions from local sources during the pollution episode were considerably higher, underscoring the need for effective management of local emission sources. Among the various emissions within Changzhou (Figure 4b), industrial processes and transportation were responsible for over 20% of ozone concentrations, followed by emissions from industrial boilers and kilns (16.1%), biogenic sources (14.7%), and power plants (13.6%).

We further examined the inter-county contributions to ozone levels within Changzhou (Figure 5a). We designated each of the seven counties as a receptor and calculated the relative ozone contributions from all other counties. Additionally, each county was treated as a source region to quantify its ozone contributions to the other counties. Taking JK as an example, Figure 5b shows the relative ozone contributions from JK and other counties to the ozone levels in JK, with the total contributions aggregating to 100%. Local emissions within JK were predominant in ozone formation, contributing up to 66.1%. WJ, located upwind from JK, exhibited the highest contribution (15.7%) among the other counties, while contributions from other districts remained below 10%. Figure 5c further illustrates the ozone contribution from sources within JK to receptors in other counties. Emissions from JK primarily impacted TN, with a contribution of 48.3%, as it was directly downwind. JK’s contributions to WJ, ZL, and XB ranged from 20% to 30%. Overall, ozone levels in ZL and XB reflected a combination of local sources and contributions from neighboring counties, whereas WJ and JK not only exhibited the highest ozone concentrations but also contributed significantly to the ozone levels in other districts (Figure A3 and Figure A4). Notably, TN was particularly vulnerable to ozone transported from adjacent counties. The uneven source–receptor ozone contributions among different counties underscored the necessity for targeted county-specific emission reduction strategies in order to achieve the most cost-effective mitigation of ozone levels for the city of Changzhou.

3.2. Effectiveness Assessment of Different Mitigation Strategies

3.2.1. Uniform Emission Reductions

The uniform emission reduction scenario applied a consistent reduction ratio (20%, 40%, and 60%) to both ANOx and AVOC emissions across Changzhou. As illustrated by Figure 6, the reduction in ANOx emissions alone resulted in a notable rebound effect on ozone concentrations. For instance, a 20% reduction in ANOx emissions during the pollution period led to an average increase of 3.7 μg·m⁻³ in the MDA8 O₃ concentrations in Changzhou, with WJ experiencing the largest increase of 5.7 μg·m⁻³. This nonlinear response of ozone concentration aligned with the CMAQ-DDM results (Figure A5 and Figure A6), which indicated a negative sensitivity of O₃ to ANOx emissions, a trend that intensified with further reductions in ANOx reductions (Figure A7). On the other hand, when AVOC emissions were reduced independently, a clear downward trend in the average MDA8 ozone concentrations was observed (Figure 7 and Figure A8). Specifically, a 20% reduction in AVOC emissions resulted in a decrease of 2.2 μg·m⁻³ in ozone levels, with JT District showing the most pronounced decline. In scenarios where both ANOx and AVOC emissions were concurrently reduced, the opposing effects of these reductions yielded a modest increase in ozone concentrations across all counties (Figure A9), with the exception of LY. Notably, LY exhibited the smallest response in ozone levels to changes in emissions, attributed to its position upwind from Changzhou, which minimized the influence of emissions from other counties (1.2% to 6.5%, as shown in Figure 5a). Furthermore, LY was unique in that it experienced reductions in ozone when ANOx and AVOC were simultaneously reduced. These findings are consistent with previous studies showing that the summertime ozone formation in Changzhou was predominantly controlled by VOC [44,45,46], necessitating an AVOC to ANOx reduction ratio greater than one to effectively mitigate ozone levels [47,48].

3.2.2. Sector-Specific Emission Reduction

Based on the CMAQ-ISAM results of sectoral contribution, the transportation and industrial processes emerged as the primary contributors to O₃ formation, and their average contributions were comparable. A second set of mitigation strategies considers a 40% emission reduction ratio applied to VOC emissions from each source sector across Changzhou. As can be seen in Figure 8a and Figure A10, VOC reductions in industrial processing led to the most substantial ozone decrease (3.2 µg·m⁻³ on average) during the pollution period, which was consistent with the ISAM results. During the last three days of the pollution period, controlling 40% of VOC emissions from industrial processing resulted in an ozone reduction of over 4.5 µg·m⁻³. Transportation appeared to be second most effective in reducing ozone concentrations. The averaged MDA8 ozone concentration decreased by 0.7 µg·m⁻³ if VOC emissions from transportation were reduced by 40%. Reductions in the other sectors resulted in minimal ozone reductions (<0.3 µg·m⁻³).

In terms of ozone changes at the county level, JT exhibited the most significant ozone reduction following emission control measures (Figure A12). When a 40% reduction in VOC emissions from industrial processes was implemented, the peak decrease in JT’s daily MDA8 O₃ concentration reached 12.7 µg·m⁻³ (3.9 µg·m⁻³ on average). Furthermore, the average MDA8 ozone concentrations in ZL decreased by 3.7 µg·m⁻³, particularly during the later stages of the pollution episode, where ozone decreased more noticeably (>6.0 µg·m⁻³). In contrast, LY experienced the least impact, with a 40% reduction in VOC emissions from various sources resulting in an average MDA8 O₃ concentration of no more than 0.5 µg·m⁻³.

3.2.3. County-Specific Emission Reduction

As demonstrated by the CMAQ-ISAM results in Section 3.1, the O₃ contributions from different counties exhibited substantial variations. These discrepancies can be attributed to the differing abundance of emissions sources within each county and the geographical positioning of each county. For instance, during the selected pollution period, particular counties (e.g., TN, ZL, XB) were significantly influenced by emissions from surrounding counties, while other counties (e.g., WJ, JK) were important contributing counties (Figure 5a). Thus, the third set of mitigation strategies focused on evaluating the effectiveness of emission reduction initiatives applied to specific counties within Changzhou. To do so, we implemented a 40% reduction in AVOC emissions for each county at a time and subsequently analyzed the corresponding changes in ozone concentrations for each case. By conducting county-specific emission reduction simulations, we provided a detailed assessment of how targeted emission reductions in specific counties can influence regional ozone levels. This approach allowed for a thorough understanding of the spatial dynamics of ozone pollution and highlights the importance of considering both local and upwind county emissions in developing effective mitigation strategies.

Figure 8b shows the changes in MDA8 O₃ concentrations resulting from AVOC reductions across different counties. On average, a 40% AVOC reduction in XB led to the most significant decrease in MDA8 O₃ concentration (0.9 µg·m⁻³) during the pollution episode, whereas the same reduction applied to LY and ZL yielded the least effective outcome (both were 0.2 µg·m⁻³). However, when evaluating daily effectiveness, no single county consistently emerged as the most effective in terms of control measures. For instance, on June 14th and 15th, AVOC reductions in JT were observed to be the most effective compared to other counties. In the following several days, reductions in WJ led to the largest ozone decrease. During the last two days of the pollution episode, reductions in XB led to the most substantial decrease in ozone concentration. The changing importance of counties to be controlled was primarily associated with the changing wind directions, as controlling emissions within the upwind counties was more effective than controlling the downwind counties. Due to the high emissions in XB, reducing XB emissions yielded the most substantial ozone reduction at regional level (Figure A11), with ozone decreases as much as 4.4 µg·m⁻³.

In terms of ozone changes at the county level, there was not always a direct correlation between the counties implementing emission reduction measures and those that experienced the resultant effects (as shown by Figure A13). For example, a 40% reduction in VOC emissions in JK resulted in minimal changes to local ozone levels, while the daily MDA8 O₃ concentration in JT decreased by 3.9 µg·m⁻³. Additionally, on 20 June, VOC reduction in ZL led to a rebound in local ozone level. In contrast, LY showed a more pronounced response to local emission reduction strategies, especially during the period from 14 to 19 June. During the last two days of the pollution episode, mitigation strategies in other counties significantly contributed to lowering ozone concentration in LY.

4. Limitations

Several limitations and uncertainties are inherent in the application of this modeling framework, which warrant brief discussion. Firstly, biases present in the WRF-simulated meteorological variables could result in either the underestimation or overestimation of the simulated ozone concentrations. The selection of physical processes and their corresponding parameterization schemes is crucial for the accuracy of meteorological field simulations [49,50]. Among these, the parameterization of the planetary boundary layer (PBL) plays a particularly important role in influencing the simulation results [51,52,53]. Chen et al. [52] employed four different PBL schemes to investigate the seasonal and diurnal variations in typical meteorological variables for the YRD region. Their findings indicated that the MYNN and MYJ schemes performed better in simulating meteorological parameters compared to the YSU and ACM2 schemes. In terms of O₃ simulation, Shi et al. [53] found that the YSU scheme was most suitable for the YRD region.

Secondly, uncertainties associated with the emission inventory would also affect model performance. In this study, the MEIC anthropogenic emission inventory was used for the D01 and D02 simulation domains. According to estimates by Zheng et al. [36], pollutants primarily emitted from large sources (e.g., SO₂) exhibit relatively low uncertainty (about ±16%). Those associate with more dispersed sources, such as black carbon (BC) and organic carbon (OC), show much higher uncertainty, ranging from −40% to 90%. For the YRD region, the estimated uncertainties in anthropogenic emissions of NOx and VOCs are ±27.7% and ±133.4%, respectively [54]. These uncertainties in anthropogenic emission inventories can directly influence the simulation of pollutant concentrations. Huang et al. [55] suggested that AVOC emissions contributed most to the uncertainty regarding predicted O₃ concentrations in coastal regions.

Thirdly, the spin-up period used in most studies is less than or equal to 10 days [55], whereas Hogrefe et al. [56] concluded that a spin-up period of 10 days or one week is insufficient to reduce the influence of initial conditions to below 1%. Our simulation started on 27 May, with a spin-up period of 5 days. To minimize the impact of initial conditions, future simulations should consider extending the spin-up period to 15–20 days.

5. Conclusions

This study provides a methodological, modeling-based framework for developing effective urban ozone pollution control strategies. Using the city of Changzhou as a case study, we demonstrate significant variations in ozone response to different mitigation strategies, highlighting the need for tailored, multi-faceted approaches that consider both sectoral and geographical factors. For sector-specific control strategy, reducing emissions from the industrial process sector yields the most substantial impact, achieving a maximum reduction of 4.2% in MDA8 O₃ concentration. For county-specific control strategy, emission reduction in XB and WJ lead to more significant decreases in ozone levels. Further simulations show that in WJ, reducing mobile emissions is more effective than reductions in other sectors, while in XB, the most significant ozone reduction comes from decreasing industrial emissions (Figure A16). However, meteorological conditions cause substantial daily fluctuations in MDA8 ozone reductions. Therefore, emission control strategies should be closely aligned with prevailing wind directions. Future research endeavors should delve deeper into ozone prevention and control strategies by integrating the emission reduction potential of different industries.

Furthermore, the ozone source apportionment analysis shows that the contribution of regional transport is equally important as that of local sources. This finding suggests that local emission reductions alone are insufficient for achieving significant air quality improvements. For instance, a 60% reduction in AVOC emissions within the city of Changzhou (referred to as Case6) resulted in an average ozone decrease of 6.5 µg·m⁻³ (maximum daily reduction of 15.7 µg·m⁻³) during the pollution episode. In contrast, a similar 60% reduction in AVOC emissions for the YRD region leads to a markedly greater reduction of 49.0 µg·m⁻³ (maximum daily reduction of 102.5 µg·m⁻³). Therefore, while it is imperative to actively promote the reduction in local pollutants, there is also a pressing need to strengthen regional cooperation and implement coordinated emission reductions at the regional level.

While the present work quantifies the ozone response to different mitigation strategies, it does not yet translate these concentration changes into public health benefits or account for the socio-economic feasibility. Extending the framework to include health-impact assessment (e.g., changes in premature mortality) would allow policymakers to weigh air-quality gains against implementation costs. Additionally, coupling the model to cost-effectiveness or cost–benefit modules, for example, considering sector-specific abatement costs, technological readiness, and regulatory enforceability, would highlight which strategies deliver the greatest health benefit per unit investment.