The WRF-Chem model (Weather Research and Forecasting model with Chemistry) (Grell et al., 2005) utilized in this study is a specific version modified by Li et al. (2010, 2011a, b, 2012) and has been extensively used for regional simulations of air pollutants in the GZB. Simulations focus on the atmospheric processes over the GZB in the warm season of 2022. This period is selected for its exceptionally high O₃ pollution (with maximum daily averaged 8 h (MDA8) O₃ concentration of 134 µg m⁻³, +7 % above the 2020–2024 mean), enabling a detailed analysis of photochemical mechanisms and co-control pathways. Figure 1 shows the model simulation domain encompassing the GZB and its surrounding regions, with particular focus on five major cities: Xi'an (XA, provincial capital), Xianyang (XY, undergoing rapid industrialization and urbanization), Weinan (WN, energy hub), Tongchuan (TC), and Baoji (BJ). Detailed model configuration can be found in Supplement Text S1 and Table S1 of the Supplement. This setup constitutes the base simulation, which serves as the reference for subsequent sensitivity scenarios.

Observations of criteria pollutants (PM_2.5, NO₂, O₃, SO₂ and CO) released by China's Ministry of Ecology and Environment (MEE) are used for characterizing recent warm-season (2014–2024) air quality trends across the GZB and validating the simulated air pollutants. Meteorological parameters, including 2 m temperature, relative humidity, and 10 m wind speed/direction, measured at Jinghe station in the GZB (34.26° N and 108.58° E) are employed to evaluate the model performance in simulating synoptic conditions. Model performance of the base simulation is evaluated against observations using statistical metrics including the mean bias (MB), root mean square error (RMSE), and the index of agreement (IOA, shown in S2 of the Supplement).

To evaluate the robustness of the simulated OFR, we have employed satellite-derived column densities of formaldehyde (HCHO) and NO₂. Daily Level-3 gridded HCHO data are obtained from the OMI/Aura HCHO Total Column Daily L3 Global 0.1° × 0.1° product (OMHCHOd v003; NASA GES DISC). Daily Level-3 gridded NO₂ data are sourced from the OMI/Aura NO₂ Cloud-Screened Total and Tropospheric Column L3 Global 0.25° × 0.25° product (OMNO2d v003; NASA GES DISC). A grid cell is excluded from the monthly average calculation if valid data are available for fewer than 8 days in that month.

Building upon the base simulation, an emission reduction matrix comprising 121 scenarios is then designed to develop EKMA diagrams to determine the O₃ and SA formation regime (Fig. S1). These scenarios include reductions of NO_X and AVOCs emissions from 0 % to 100 % with an interval of 10 %. Given that the biogenic VOCs (BVOCs) emissions are uncontrollable, only the AVOCs emission reduction is taken into consideration. It does not mean the insignificance of BVOCs in the O₃ formation. However, changes of plant cover and emissions as well as the O₃ uptake capacity of plants all affect the near-surface O₃ concentration through ecosystem-atmosphere interactions (Lin et al., 2020). The high-resolution, online-coupled WRF-Chem framework simulates all key processes affecting O₃ formation (local photochemistry, regional transport, vertical mixing) consistently. By varying emissions under fixed meteorology and boundary conditions, the model isolates the net impact of emission changes on MDA8 O₃ concentrations. While the classical EKMA method based on O₃ production rates reflects a more localized chemical mechanism, our approach provides an integrated, policy-relevant sensitivity diagnosis that accounts for both formation and transport at the urban-regional scale, offering directly actionable insight for control strategies (Ye et al., 2025). Furthermore, the brute force method (BFM) is used to evaluate the O₃ and SA contributions of industry, power, residential, transportation and biogenic sources in the GZB and five cities during the warm season of 2022. The BFM calculates the contribution of a specific source sector by completely removing (setting to zero) all emissions from that sector in the base simulation and comparing the resulting MDA8 O₃ or SA concentration with the base case (Dunker et al., 1996). In the present study, MDA8 O₃ concentrations are used as the response metric in chemical regime identification and source apportionment for several key reasons. First, MDA8 O₃ represents the period of highest daily O₃ exposure, offering a robust indicator of actual O₃ pollution levels. Second, it is the official metric for ambient O₃ standards under China's national air quality regulations. Assessing how MDA8 O₃ responds to emission perturbations therefore provides direct, policy-relevant insights for designing effective air quality management and control strategies.

To understand the evolving characteristics of summertime photochemical pollution in the GZB, we first examine recent trends in near-surface O₃ and PM_2.5 concentrations during the warm seasons from 2022 to 2024. This period represents the most recent stage of air quality management in the region and provides insight into ongoing challenges posed by secondary pollutants. Under China's sequential air pollution policies – the Air Pollution Prevention and Control Action Plan (2013–2017; State Council, 2013), the Three-Year Action Plan to Win the Blue-Sky War (2018–2020; State Council, 2018), and the Air Quality Continuous Improvement Action Plan (2023–2025; MEE, 2023) – the GZB has achieved sustained PM_2.5 reductions. Warm-season PM_2.5 concentrations have decreased from approximately 43.1 µg m⁻³ in 2014 to around 21.6 µg m⁻³ in 2024, representing a nearly 50 % reduction (Fig. 2a). In contrast, the mean MDA8 O₃ concentration during the warm season has increased from 96.0 µg m⁻³ in 2014 to roughly 132.7 µg m⁻³ in 2024, with a substantial increase from 2014 to 2017, a decrease from 2017 to 2020, followed by a renewed upward trend thereafter (Fig. 2b). While the substantial alleviation in the particulate pollution is mainly due to the anthropogenic emission mitigation, O₃ pollution has worsened in the region primarily driven by increased unfavorable conditions and secondarily by changes in anthropogenic emissions (Bei et al., 2022). Near-surface observations have revealed persistent O₃ pollution during the warm season, with frequent exceedances of the national ambient air quality standard for MDA8 O₃ (160 µg m⁻³, as shown in Table 1). Year-to-year fluctuations in mean MDA8 O₃ levels are governed by the frequency of exceedances, which rise from 8.3 % in 2014 to 36.7 % in 2017, fall to 13.6 % in 2020, and climb again to 27.0 % in 2024; the magnitude of those exceedances remains relatively stable, ranging from 178 to 189 µg m⁻³ (Table 1).

The monthly evolution of O₃ concentrations reveals a pronounced sub-seasonal pattern in the GZB. June consistently emerges as the peak month for both MDA8 O₃ levels and exceedance frequency (Fig. S5 and Table S2). During the warm seasons from 2014 to 2024, the mean MDA8 O₃ concentration in June reaches approximately 132 µg m⁻³, with over 30 % of days with O₃ exceedances. The peak corresponds closely with meteorological conditions typically observed in early summer, when the region experiences strong solar radiation, elevated temperatures, and relatively low precipitation conditions that are highly conducive to photochemical O₃ formation (Pay et al., 2019). The early-summer O₃ peak observed in the GZB is consistent with reports from northern China, where June maxima are evident in the Beijing–Tianjin–Hebei (BTH) region and June–July peaks prevail across the North China Plain (Han et al., 2020; Li et al., 2020). In contrast, in southern China, O₃ sub-seasonality is strongly modulated by subtropical high-pressure systems, the East Asian monsoon, typhoon passages, and land–sea breeze circulations. As a result, the peak O₃ episodes are tended to emerge in May or during the late summer to early autumn (September–October) in the Yangtze River Delta (YRD) and Pearl River Delta (PRD) (Han et al., 2020; Xu et al., 2020; Ouyang et al., 2022). These regional contrasts highlight that while the precise timing of seasonal O₃ peaks is shaped by local climate and meteorology, the emergence of a distinct sub-seasonal maximum is a robust feature across China's major urban clusters. This underscores the sensitivity of regional air quality to meteorological transitions and highlights the need for sub-seasonally adaptive control strategies, in particularly during the high-risk period.

How these basin-wide trends manifest at the city scale is then explored. O₃ variations in all five cities mirror the overall pattern in GZB, with MDA8 O₃ concentrations rising from 2014 to a maximum in 2017 and then dipping through 2020 before climbing again to 2024 levels. Cities of XA, XY, and WN, located in the central GZB, have the relatively high O₃ levels, with the mean MDA8 O₃ concentration ranging of 128–129 µg m⁻³ during the warm seasons of 2014–2024. XY has experienced the fastest MDA8 O₃ concentration increase of 6.3 µg m⁻³ yr⁻¹, nearly twice the basin average of 3.7 µg m⁻³ yr⁻¹, which likely reflects rapid urban expansion and increasing local precursor emissions. The city also records the highest fraction of exceedance days, averaging 28.7 % of warm-season days during 2014–2024 and peaking at 49.2 % in 2017 (Table 1). XA is also characterized by the sever O₃ pollution comparing with other cities, with a growth rate of 3.8 µg m⁻³ yr⁻¹ and exceedances on 25.9 % of the warm-season days, consistent with its role as the region's primary emission hub. By contrast, BJ, situated on the western edge and often upwind of the basin core, had the lowest O₃ burden, with the mean MDA8 O₃ concentration of 113.3 µg m⁻³ and only 10.9 % exceedances of warm-season days during 2014–2024. These spatial differences underscore the importance of both local precursor controls in rapidly urbanizing cities and regional transport pathways in shaping O₃ pollution across the GZB.

Generally, the long-term observations reveal not only a basin-wide deterioration in O₃ pollution but also strong sub-seasonal and spatial heterogeneity, highlighting the necessity of process-based modeling to resolve spatiotemporal patterns in warm-season O₃ formation characteristics and to provide insights into air pollution mitigation in the GZB.

The meteorological simulations during the warm season of 2022 demonstrate excellent reproduction of diurnal temperature patterns (IOA = 0.99) despite slight overestimation biases (+0.4 °C), while relative humidity variations are well captured (IOA = 0.95) with marginal underestimation (-1.8 %, Fig. S2). Spatially, the model reproduces key air pollutants distributions across the GZB against measurements (Fig. S3). Elevated PM_2.5 and O₃ in the eastern and central regions arise from northeasterly transport, weak winds over the central basin plain that favor accumulation of air pollutants, and southerly flows carrying BVOCs from the Qinling forests. Temporally, simulated air pollutant concentrations show good agreement with observations with IOAs all exceeding 0.5 (Fig. S4). The model's good performance in replicating synoptic-scale meteorological processes and associated air pollutants warrants its suitability for mechanistic analysis. Comprehensive validation of the WRF-Chem model performance is detailed in S3 of the Supplement.

O₃ formation in the planetary boundary layer (PBL) is a complex and nonlinear process driven by sunlight acting on NO_X and VOCs. Figures 3 and 4 present EKMA diagrams for four high-O₃ pollution episodes from May to August 2022 in urban areas of the GZB and its five cities, respectively. These diagrams depict O₃ isopleths for OFR identification, derived from sensitivity simulations with systematically reduced NO_X and AVOCs emissions. The ridge line (red lines) delineates the boundary between these regimes: scenarios above it lie in the VOCs-limited regime (O₃ falls more with AVOCs cut), those below in the NO_X-limited regimes, and scenarios near the line are transitional regimes (mixed sensitivity). The upper-right corner (100 % AVOCs, 100 % NO_X emissions) represents the current emission scenario, whose location relative to the ridge line determines the prevailing sensitivity regime.

Quantifying sectoral drivers of warm-season O₃ production bridges the sensitivity diagnostics from Sect. 3.2 with operational emission control design. The MDA8 O₃ contributions of industrial, residential, traffic, power plants and biogenic sources are assessed from May to August using the BFM (Figs. 7 and 8).

The attribution results show pronounced spatiotemporal heterogeneous across the GZB. At the basin scale, the contribution of all anthropogenic sources increases from 32.8 % in May to 55.2 % in July, then declines to 48.0 % in August (Fig. 7). The rise is largely driven by increasing industrial and traffic influence: industrial emissions contribute 5.4 % to MDA8 O₃ in May, rising to 11.8 % in July, while traffic contributions increase from 5.8 % to 17.0 % during the same period (Fig. 8 and Table 2). Because NO_X and AVOCs emissions peak in June and then decline, the continued anthropogenic contribution growth through July indicates that enhanced photochemical activity under strong solar radiation and rising BVOCs emissions further amplify O₃ formation. Biogenic contributions notably rise from 9. 7 % in May to 18.7 % in July then slightly decline to 16.7 % in August (Fig. 7 and Table 2), consistent with elevated BVOCs emissions under warmer, sunnier conditions. These sub-seasonal dynamics are consistent with findings from other Chinese urban clusters, such as the BTH, YRD and PRD, where rising BVOCs emissions in summer have been shown to enhance photochemical reactivity and partially offset gains from anthropogenic VOCs and NO_X reductions, thereby promoting shifts of OFR toward NO_X-limited or transitional regimes (Wu et al., 2020b; Zhao et al., 2022; Gao et al., 2025; Wang et al., 2025a).

Spatial contrasts across the five cities further illustrate how emission profiles interact with chemical regimes and guide targeted mitigation polices. XA and XY show the largest anthropogenic O₃ shares, rising from 37.1 % in May to 58.7 % in July in XA, and from 35.7 % to 59.2 % in XY (Fig. 7 and Table 2). In these urban cores, industrial and traffic emissions are dominant, contributing approximately 9 %–12 % (May) and 26 %–31 % (June-August) to MDA8 O₃ concentrations (Fig. 8). TC and BJ show intermediate anthropogenic contributions. Traffic dominates their O₃ production (7.1 %–15.9 % in TC and 7.9 %–18.3 % in BJ), and industrial emissions are the secondary contributor to the O₃ levels. Given their VOCs-limited regime (Sect. 3.2) and higher industrial AVOCs emissions (Table S5) in May–June, prioritizing reductions in industrial AVOCs is advisable to prevent O₃ rebound. As the OFR shifts toward transition and NO_X-limited conditions in late summer, mitigation efforts should be combined with traffic emissions due to the higher NO_X share from vehicular exhaust. In industrial regions, for instance, Dai et al. (2025) have showed that oxygenated VOCs (OVOCs) contribute a high proportion (∼ 30 %–37.8 %) of VOCs pools in industrial cities, enhancing radical production and O₃ sensitivity; thus, control strategies must consider VOCs speciation, not just total amounts. By contrast, WN exhibits much lower anthropogenic influence (13.9 %–34.7 %, Fig. 7 and Table 2), with power plant emissions exerting a net consuming effect on local O₃. This is attributed to strong local NO titration and enhanced termination pathways reduce OH / HO₂ recycling. Similar suppression effects have been reported in Zhao et al. (2025) that power plant NO_X emissions contribute negative O₃ signals in certain regions, indicative of O₃ titration under saturating NO_X. Meanwhile, analyses in the YRD have showed that during emission reductions, weakened NO titration can drive O₃ increases (∼ 20 % of the rise attributed to reduced NO titration) (Wang et al., 2022b). Mechanistic modeling also indicates that NO_X reductions in VOCs-limited areas may lead to O₃ increases, unless VOCs controls are pursued simultaneously (Dai et al., 2024a; Tang et al., 2021). Therefore, aggressive NO_X cuts alone risk raising O₃ unless paired with targeted AVOCs reductions in WN.

These variations highlight that a uniform, time-invariant control strategy is insufficient. Instead, effective O₃ mitigation in the GZB demands sub-seasonally adaptive, sector-specific emission controls. In early summer, dominant VOCs sensitivity suggests focusing on industrial AVOCs; in mid to late summer, integrating AVOCs and traffic NO_X controls aligns better with evolving regimes; and in regions like WN, VOCs-targeted strategies must accompany any NO_X reductions. Recent studies similarly argue that tailoring precursor reductions to local O₃ regimes yields greater benefits than uniform cuts. For example, Zhu et al. (2022) showed that in the NCP and YRD, formation regimes have shifted from VOCs-limited toward transitional or NO_X-limited states as atmospheric oxidation increases. Wang et al. (2025b) reported pronounced vertical and spatial differences in OFRs in eastern Chinese cities, implying that a one-size-fits-all control is often suboptimal. Li et al. (2024) have used FNR (HCHO / NO₂) diagnostics at multiple altitudes to reveal that optimal precursor strategies vary by vertical layer in the BTH.

The source contributions presented in this study are quantified using the BFM, wherein emissions from a specific sector are entirely removed to evaluate its potential impact on O₃ concentrations. While this approach is effective for assessing the control potential of individual sources and is widely used in policy-relevant scenario analysis, it is important to consider the inherent nonlinearity of tropospheric O₃ chemistry. The response of O₃ concentrations to a given emission change depends strongly on the background chemical regime (NO_X- vs. VOCs-limited, as shown in Sect. 3.3). For example, in a VOCs-limited regime, reducing NO_X emissions alone may inadvertently increase O₃ concentrations, whereas the same reduction in a NO_X-limited regime would lower O₃ concentrations. Consequently, the attribution derived from complete removal of a source may not scale linearly with incremental emission controls. As discussed in Li et al. (2023), when emission perturbations are large, the assumption of a linear response between emission change and O₃ change becomes invalid due to the nonlinear interactions in O₃ production chemistry, meaning that attribution results from BFM may differ from those obtained by alternative methods such as source tagging that explicitly follow chemical pathways. Importantly, BFM and tagging methods answer different scientific questions that the BFM estimates the sensitivity of O₃ to sector-specific emission changes, whereas tagging method attributes the total mass of O₃ to source precursors independent of changes in emissions (Li et al., 2023; Shu et al., 2023). In this study, the BFM results are interpreted in conjunction with the OFR analysis (Sect. 3.3), which delineates the nonlinear photochemical sensitivity to gradual precursor changes. Together, they provide a robust basis for identifying priority source sectors and designing effective, regime-specific control strategies for the GZB.