Near Real-Time Distribution of Ozone in China from 2013 to 2020 and Agricultural Impacts
- 1Ministry of Education Key Laboratory for Earth System Modeling, Department of Earth System Science, Tsinghua University, Beijing, China (liujk19@mails.tsinghua.edu.cn)
- 2State Key Joint Laboratory of Environment Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing, China (guannangeng@tsinghua.edu.cn)
- 3State Key Laboratory of Severe Weather & Key Laboratory of Atmospheric Chemistry of CMA, Chinese Academy of Meteorological Sciences, Beijing, China (zhongjt@cma.gov.cn)
Air pollution is one of the most important environmental problems in China. As a major air pollutant, ozone (O3) will endanger human health and terrestrial ecosystems. It is of great practical significance to obtain a continuous full-coverage dataset of ozone with high spatio-temporal resolution to conduct mechanism research from its causes, development, diffusion, impact and other aspects. In this study, a 3-stage machine learning model was developed through multiple data fusion, and the LightGBM method is used to obtain the hourly spatio-temporal distribution dataset of the O3 concentration in China from 2013 to 2020, with a resolution of 0.25 °× 0.25°. We first revise the meteorological reanalysis data using ground observation and propose a data fusion algorithm to achieve the ground level distribution of ozone, which combines ground observation of pollutants, population data, revised reanalysis meteorological conditions, reanalysis of radiation, land and vegetation data, emission inventory and results of chemical transport model simulation. In addition, due to the common phenomenon that the previous prediction models underestimate the extreme value of the pollution periods,therefore, we redefined the heavy pollution event and assimilated it into the 0.25 grid by using the synthetic minimum oversampling technique (SMOTE) method to improve the model performance during the extreme pollution periods.
Our model, with the 10-fold CV result of R2 = 71% and RMSE= 25.1μg·m-3, and our hourly O3 concentration results are spatially and spatially continuous with a similar distribution compare to the observation, which proves the reliability of our model. With higher time resolution, various exposure response indicators can be obtained. AOT40 calculated by high-resolution hourly ozone concentration further, which is far more accurate than it when directly predicted by daily indexs modeling.
In addition, based on the distribution of AOT40, we assessed the agricultural damage and ecological damage caused by the change of surface ozone pollution during 2013-2020. Our estimation considered the planting area and phenological period of crops that the overestimation of crop RYL in the region can be avoided. The annual avrage production loss of wheat, rice and maize in China from 2013 to 2020 is 55.0, 57.4 and 23.6 Mt, respectively. Besides, The loss of gross primary productivity was also estimated. During 2013-2020, the ozone pollution in China caused an annual average loss of 2.1%, and the loss in the south was much higher than that in the north.
How to cite: Liu, L., Geng, G., Zhong, J., Liu, Y., Xiao, Q., Zhang, X., and Zhang, Q.: Near Real-Time Distribution of Ozone in China from 2013 to 2020 and Agricultural Impacts, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-6562, https://doi.org/10.5194/egusphere-egu23-6562, 2023.
