Potential Biases and Driving Factors in the Estimation of China's Background Ozone: A Comparative Study of Observation-, Model-, and Fusion-Based Methods

Tropospheric background ozone (O3) is a critical benchmark for refining ambient air quality standards and assessing emission control potential. However, large discrepancies exist among different estimation methods, limiting the reliability of policy-relevant insights. Here, we compared estimations of China background O3 (CNB O3) using three representative methods: Principal Component Analysis (PCA, observation-based), Chemical Transport Model with Brute-Force Method (CTM-BFM, model-based), and Model-Measurement Fusion (MMF, fusion-based). An ensemble framework was further applied to derive robust CNB O3 estimates, while methodological uncertainties and underlying drivers were systematically investigated. Results show that the 2018–2020 ensemble mean (EM) CNB MDA8 O3 was 37.6 ppb ± 4.7 ppb at the national scale, with distinct spatiotemporal patterns shaped by natural photochemical processes (dominated by surface temperature, 39%, and relative humidity, 12%), transboundary transport (indicated by latitude, 26%, and longitude, 5%), and stratosphere-troposphere exchange (indicated by altitude, 7%). Methodological uncertainty, quantified as the relative standard deviation (SD%) among the three methods, averaged 13-14% nationally with substantial regional heterogeneity: SD%≤15% in Western China versus >25% in Eastern China during summer. Machine learning-based interpretation further revealed divergent representations of photochemical reactions and transboundary transport as the primary sources of discrepancy between PCA and CTM-BFM. MMF achieved the closest agreement with EM by integrating observational constraints and physical mechanisms, despite higher computational costs. This integrated ensemble analysis and systematic uncertainty quantification provide robust scientific support for O3 pollution mitigation and targeted air quality management in China.