Insights from ozone and particulate matter pollution control in New York City applied to Beijing

Elevated concentrations of ozone (O3) and/or fine particulate matter (PM2.5) negatively affect human health and are a major concern in densely populated megacities. O3 and PM2.5 frequently co-occur during summer due to stagnant meteorology with high solar radiation and temperature, under which elevated nitrogen oxides (NOx) and volatile organic compounds (VOCs) promote formation of both pollutants. Prior work has shown a positive relationship between maximum daily 8-hour average (MDA8) O3 and daily 24-hour average (DA24) PM2.5 in polluted regions during summer, with a maximum turning point (MTP) around 50–60 µg m−3 PM2.5 in Chinese megacity clusters, beyond which O3 remains relatively stable due to radical uptake by particles inhibiting O3 production. NYC and Beijing have implemented strict emission controls (NYC since the 1970s, Beijing since 2013), resulting in substantial PM2.5 decreases, while O3 decreased more slowly in NYC and increased in Beijing. This discrepancy motivates examining how emission controls altered the O3–PM2.5 relationship. The study analyzes multi-year surface measurements (NYC: 19 summers; Beijing: 6 summers) and aerosol composition, fits a non-linear function capturing co-occurrence and PM2.5-induced O3 suppression, and uses these relationships to guide modeled control strategies to jointly manage PM2.5 and O3 in Beijing and other Chinese megacity clusters.

The study combined observational analysis, empirical non-linear fitting, aerosol composition measurements, and 3-D chemical transport modeling. - Non-linear O3–PM2.5 relationship fitting (Method M1): A simple empirical function was used to represent two effects: (a) a positive linear term for co-occurrence driven by shared precursors (NOx, VOCs); (b) a negative power function term for PM2.5 suppression of O3 formation via uptake of radicals (HO2/NO2) and reduced photolysis with higher aerosol loading. The function was O3 = α·PM2.5 + β·(PM2.5)^(5/3) + c, where α is the linear slope, β (< 0) represents suppression strength, and c is a constant. The 5/3 exponent is based on aerosol surface area scaling (~PM2.5^(2/3)) affecting radical uptake and radical levels scaling with PM2.5. PM2.5 data were binned (typically 5 µg m−3 bins) and non-linear least-squares fitting was applied. Sensitivity checks (e.g., 2 µg m−3 bins) were performed. - Aerosol chemical composition (Method M2): Non-refractory PM1 composition was measured with Aerodyne Aerosol Mass Spectrometer (AMS) in representative summers and assumed similar to PM2.5 composition based on prior evidence that PM1 dominates PM2.5 and component fractions are similar. Primary black carbon was not included (AMS limitation). Composition was grouped into: SAP (sulfate, ammonium from ammonium sulfate, and primary organic aerosol) which do not share NOx/VOC precursors with O3; and non-SAP (secondary organic aerosol, nitrate, and nitrate-related ammonium) formed from NOx/VOCs. For NYC, AMS data from summers 2001, 2009, 2011, 2018 represented subperiods SPNY1 (2001–2003), SPNY2 (2004–2008), SPNY3 (2009–2013), SPNY4 (2014–2019). For Beijing, AMS data from June–July 2014 and June 2017 represented SPBJ1 (2014–2016) and SPBJ2 (2017–2019). Uncertainties arise from using PM1 as proxy for PM2.5, omission of black carbon, and single-year representation of subperiods. - CMAQ modeling (Method M3): CMAQ v5.2 with SAPRC-07 and updated AERO6 was used over East Asia (36 km × 36 km grid) for June–August 2017. Heterogeneous losses of NO2, SO2, glyoxal, methylglyoxal and HO2 uptake on aerosols were included; HO2 uptake coefficient γHO2 = 0.2 was assumed (noting dependence on composition, T, RH and associated uncertainty). Meteorology from WRF v4.0 with 3D nudging aloft; emissions: MEIC v1.3 (China, 2016 annual used for 2017 summer) and REAS v3.1 (rest of domain) at 0.25° resolution; biogenics from MEGAN v2.1; fires from FINN; dust and sea salt inline; no lightning NOx. Initial/background conditions used CMAQ defaults (clean continental, O3 30–70 ppb). Three-day spin-up. Model skill was evaluated against observations for 24 cities in BTH region and met recommended benchmarks. Scenario cases applied proportional reductions of all anthropogenic emissions by 25%, 50%, 75%. - Estimating synchronous (equal-percentage) emission reductions (Method M4): Relationships between top-5% DA24 PM2.5 and emission reduction ratio from CMAQ were used to infer required regional equal-percentage reductions to achieve targets. Model-observation scaling factors were applied (Beijing: 1.2; Shanghai: 1.0; Guangzhou: 0.9) to correct for model bias. Targets: Goal 1 (NYC 2001 top-5% PM2.5: 39 µg m−3), Goal 2 (O3 standard 75 ppb mapped to region-specific PM2.5 via fitted O3–PM2.5 curves), Goal 3 (NYC 2019 DA24 PM2.5: 15 µg m−3).

- NYC: The O3–PM2.5 linear slope increased from ~2.0 (SPNY1/2) to ~3.86 (SPNY4), indicating weaker O3 control relative to PM2.5 after emission controls. Extreme summer concentrations (top 5%) declined 2001–2019 at 1.1 ppb yr−1 for MDA8 O3 and 1.9 µg m−3 yr−1 for DA24 PM2.5, totaling −22% O3 and −62% PM2.5. SAP mass fraction in PM decreased from 51% (SPNY1) to 29% (SPNY4); adjusting for SAP fraction reduces inter-period slope differences, implicating composition change as key driver. The fraction of (NOx + VOCs) in total emissions (NOx, VOCs, SO2, PM2.5) increased from 69% (2001) to 85% (2017), favoring non-SAP and O3 formation and elevating O3–PM2.5 slopes. The suppression strength (power function coefficient β magnitude) increased from 0.08 (SPNY1) to 0.23 (SPNY4), implying enhanced PM2.5-induced O3 suppression per unit PM2.5 at later years, though mechanism details are uncertain. - Beijing/BTH: Despite strong reductions in SO2 and primary PM2.5 since 2013, summertime O3 remained high. 2016 vs 2014 emission changes: NOx −12%, VOCs ~0%, PM2.5 −28%, SO2 −39%. The O3–PM2.5 linear slope increased from SPBJ1 (2014–2016) to SPBJ2 (2017–2019). The suppression coefficient increased from 0.02 to 0.05. Fitted MTP shifted from ~140 µg m−3 (SPBJ1) to ~83 µg m−3 (SPBJ2), though the MTP magnitude/location depends on the simple fit and is uncertain. Adjusting for SAP fraction changes yields similar O3–PM2.5 relationships across subperiods, again pointing to composition/emission mix as the driver. These relationships extend regionally across BTH urban areas. - Modeled control strategy (regional equal-percentage reductions) for BTH (based on 2019 emissions): estimated regional reductions required: 42% to reach Goal 1 (Beijing top-5% PM2.5 down to 39 µg m−3), 53% for Goal 2 (MDA8 O3 < 75 ppb, corresponding to PM2.5 ≈ 29 µg m−3 in Beijing), and 70% for Goal 3 (DA24 PM2.5 = 15 µg m−3). Keeping the emission mix constant avoids further increases in the O3–PM2.5 slope and allows simultaneous reductions of both pollutants. - YRD and PRD: YRD’s O3–PM2.5 slope increased from SP1 (2014–2016) to SP2 (2017–2019), consistent with larger SO2 and primary PM2.5 reductions relative to NOx/VOCs; suppression effect increased. Estimated YRD regional reductions (based on Shanghai 2019 and YRD emissions) to meet goals: 28% (Goal 1), 32% (Goal 2), 59% (Goal 3). PRD showed only slight slope increase and similar suppression across subperiods due to smaller changes in SO2/PM2.5 vs NOx/VOCs; Guangzhou 2019 top-5% PM2.5 (36.2 µg m−3) already below Goal 1, but additional reductions of 14% (Goal 2) and 69% (Goal 3) are needed. - Indicative linkage to O3–NOx–VOC sensitivity: Across regions, fitted suppression coefficients increased in ways consistent with regime transitions: BTH 0.02→0.06 (VOC-limited toward transitional), YRD 0.02→0.07 (similar shift), PRD 0.06→0.08 (staying transitional/weak-VOC-limited), NYC 0.08→0.23 (toward transitional/weak-NOx-limited).

Current control strategies emphasizing SO2 and primary PM2.5 reductions more than VOCs/NOx have altered PM2.5 composition (reduced SAP fraction) and increased the O3–PM2.5 linear slope, meaning O3 decreases more slowly than PM2.5. In Beijing’s VOC-limited conditions (2014–2016), greater NOx reductions than VOCs further contributed to slope increases. Empirically and in CMAQ simulations, applying regional equal-percentage reductions across PM2.5, SO2, NOx, and VOCs holds the emission mix constant, suppresses further growth of the O3–PM2.5 slope, and enables concurrent reductions in both pollutants. The fitted power-function suppression term suggests PM2.5 can damp O3 formation, with its magnitude potentially tied to aerosol composition and the prevailing O3–NOx–VOC sensitivity regime. While these findings inform control design, equal-percentage reduction is one of many possible strategies; optimized combinations of NOx and VOC controls tailored to local chemical regimes could perform better. Lessons from New York State’s large NOx (−66%) and VOC (−65%) reductions since 2001 highlight the importance of substantial, balanced precursor controls from vehicles, energy generation, industrial processes, solvent use, and combustion.

The study demonstrates that emission controls in NYC and Beijing have reshaped the summertime O3–PM2.5 relationship: reductions in SO2 and primary PM2.5 relative to NOx/VOCs lowered the SAP fraction, increased the O3–PM2.5 linear slope, and enhanced PM2.5’s suppression of O3 formation. Empirical fits and CMAQ simulations indicate that maintaining a constant emission mix via regional equal-percentage reductions can avoid further slope increases and enable simultaneous decreases in O3 and PM2.5. Estimated regional reduction requirements were quantified for BTH/Beijing and extended to YRD/Shanghai and PRD/Guangzhou for multiple air-quality goals. The work also suggests a possible linkage between the fitted suppression coefficient and O3–NOx–VOC sensitivity regime. Future work should: (1) test a broader suite of NOx/VOC control combinations; (2) develop more mechanistic, accurate non-linear functions; (3) confirm the existence and location of MTP with expanded observations and modeling; and (4) reduce uncertainties in aerosol composition effects and heterogeneous radical uptake parameters.

- Empirical non-linear fit is simplified; the chosen 5/3 exponent and functional form may not fully capture mechanisms, introducing uncertainty in slope, suppression coefficient, and MTP estimates. - The existence and precise location of the maximum turning point (MTP) are sensitive to the fitting approach and available data; more observations and modeling are needed for confirmation. - Aerosol composition inferred from PM1 AMS measurements was assumed representative of PM2.5; primary black carbon was not measured by AMS and was neglected. Single-year AMS datasets represented multi-year subperiods, introducing uncertainty. - Sparse detailed measurements limit mechanistic attribution for the observed increase in suppression coefficient (e.g., deposition/uptake processes). - CMAQ configuration uncertainties: assumed constant HO2 uptake coefficient (γHO2 = 0.2) despite dependence on composition, temperature, and RH; no lightning NOx; background conditions and spin-up choices may affect results. - Emission inventories (MEIC/REAS) and their temporal allocation carry uncertainties; 2016 emissions were used to represent 2017 summertime. - Equal-percentage reduction scenario is illustrative and may not be optimal; region-specific chemical regimes may require tailored NOx/VOC control ratios.