A machine learning approach to quantify meteorological drivers of ozone pollution in China from 2015 to 2019

Abstract

Abstract. Surface ozone concentrations increased in many regions of China from 2015 to 2019. While the central role of meteorology in modulating ozone pollution is widely acknowledged, its quantitative contribution remains highly uncertain. Here, we use a data-driven machine learning approach to assess the impacts of meteorology on surface ozone variations in China for the period 2015–2019, considering the months of highest ozone pollution from April to October. To quantify the importance of various meteorological driver variables, we apply nonlinear random forest regression (RFR) and linear ridge regression (RR) to learn about the relationship between meteorological variability and surface ozone in China, and contrast the results to those obtained with the widely used multiple linear regression (MLR) and stepwise MLR. We show that RFR outperforms the three linear methods when predicting ozone using local meteorological predictor variables, as evident from its higher coefficients of determination (R2) with observations (0.5–0.6 across China) when compared to the linear methods (typically R2 = 0.4–0.5). This refers to the importance of nonlinear relationships between local meteorological factors and ozone, which are not captured by linear regression algorithms. In addition, we find that including nonlocal meteorological predictors can further improve the modelling skill of RR, particularly for southern China where the averaged R2 increases from 0.47 to 0.6. Moreover, this improved RR shows a higher averaged meteorological contribution to the increased trend of ozone pollution in that region, pointing towards an elevated importance of large-scale meteorological phenomena for ozone pollution in southern China. Overall, RFR and RR are in close agreement concerning the leading meteorological drivers behind regional ozone pollution. In line with expectations, our analysis underlines that hot and dry weather conditions with high sunlight intensity are strongly related to high ozone pollution across China, thus further validating our novel approach. In contrast to previous studies, we also highlight surface solar radiation as a key meteorological variable to be considered in future analyses. By comparing our meteorology based predictions with observed ozone values between 2015 and 2019, we estimate that almost half of the 2015–2019 ozone trends across China might have been caused by meteorological variability. These insights are of particular importance given possible increases in the frequency and intensity of weather extremes such as heatwaves under climate change.

Shifts of Formation Regimes and Increases of Atmospheric Oxidation Led to Ozone Increase in North China Plain and Yangtze River Delta From 2016 to 2019

“…During simulation periods, emissions of HCHO and other VOCs have been declined due to the strict clean air policies (Gu et al., 2020), which may lead to the lower levels of HCHO. However, the enhanced AOC associated with the emission reduction (Y. X. Liu et al., 2020; S. Q. Zhu et al., 2021) and unfavorable meteorology (Weng et al., 2022) tended to promote the secondary formation of HCHO, which offset the impact of emission reduction in HCHO. Moreover, our source apportionment of HCHO showed the contribution of industry increased 44% (from 0.25 to 0.36 ppb) from 2016 to 2019 in NCP region, while the source apportionment of NO 2 changed little in NCP and YRD regions (Figures S16 and S17 in Supprting Information ).…”

“…In 2019, the annual levels of OH and HO 2 radicals were up to ∼0.07 and ∼6 ppt in NCP regions, respectively (Figure 5), which were comparable with the previous studies (Ma et al, 2019) (0.07 ± 0.35 and 0.053 ± 0.32 ppt for OH radicals during the clean and polluted episodes, respectively in NCP). The obvious elevation of OH and HO 2 in 2019 could be explained by the unfavorable meteorological conditions (Weng et al, 2022) and the decline of NO x , which was the primary HO x (OH and HO 2 ) sink. Importantly, AOC defined as the sum of simulated oxidation rates of VOCs via reactions with HO x (OH and HO 2 ) and NO 3 radicals showed the similar results, with the highest value up to 6 × 10 −42 mol/(cm 3 • s) in the summer (Figure S27 in Supporting Information S1).…”

Deep Learning‐Based Ensemble Forecasts and Predictability Assessments for Surface Ozone Pollution

“…These simulations produced 16,464 hourly WRF-GC outputs, which we randomly divided into a training set (85%, 13,994 hr) and a test set (15%, 2470 hr). We further conducted a "WG-FNL-2019" simulation for an independent year (2019) and used the results (2,496 hr) to validate the 2DCNNs (Weng et al, 2022). Finally, we drove and nudged the WRF-GC model with 10 members from the ensemble data assimilation (EDA) (Isaksen et al, 2010) of the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ECMWF ERA5 https:// www.ecmwf.int/en/forecasts/dataset/ecmwf-reanalysis-v5, last accessed: 6 May 2022; Hersbach et al, 2020) for 2021.…”

O3–precursor relationship over multiple patterns of timescale: a case study in Zibo, Shandong Province, China

“…As the timescale changed from a wider (i.e., 5-month scale) to narrower (i.e., weekly scale) pattern, the uncertainty (indicated by the average, median, and 25 %-75 % quantile) decreased accordingly. In addition, meteorological factors such as temperature and irradiation also play an important role in O 3 formation, and it is specifically noted that these meteorological parameters can vary greatly over a long observational period (Boleti et al, 2020;Liu et al, 2019;Weng et al, 2022). Therefore, the masked temporal variation of these meteorological factors behind the averaged input dataset would also result in model uncertainty.…”