Rebuilding high-quality near-surface ozone data based on the combination of WRF-Chem model with a machine learning method to better estimate its impact on crop yields in the Beijing-Tianjin-Hebei region from 2014 to 2019

Tian Han; Xiaomin Hu; Jing Zhang; Wenhao Xue; Yunfei Che; Xiaoqing Deng; Lihua Zhou

doi:10.1016/j.envpol.2023.122334

Rebuilding high-quality near-surface ozone data based on the combination of WRF-Chem model with a machine learning method to better estimate its impact on crop yields in the Beijing-Tianjin-Hebei region from 2014 to 2019

Environ Pollut. 2023 Nov 1:336:122334. doi: 10.1016/j.envpol.2023.122334. Epub 2023 Aug 9.

Authors

Tian Han¹, Xiaomin Hu¹, Jing Zhang², Wenhao Xue³, Yunfei Che⁴, Xiaoqing Deng¹, Lihua Zhou⁵

Affiliations

¹ Faculty of Geographical Science, Beijing Normal University, Beijing, 100875, China.
² Faculty of Geographical Science, Beijing Normal University, Beijing, 100875, China. Electronic address: jingzhang@bnu.edu.cn.
³ School of Economics, Qingdao University, Qingdao, 266071, China.
⁴ Key Laboratory for Cloud Physics of China Meteorological Administration (CMA), CMA Weather Modiﬁcation Centre, Beijing, 100081, China.
⁵ Department of Earth System Science, Tsinghua University, Beijing, 100084, China.

PMID: 37567405
DOI: 10.1016/j.envpol.2023.122334

Abstract

In recent years, the problem of surface ozone pollution in China has been of great concern. According to observation data from monitoring stations, the concentration of near-surface ozone (O₃) in China has gradually increased in recent years, and ozone concentration often exceeds the contaminant limit standard, especially in the Beijing-Tianjin-Hebei (BTH) region. High O₃ concentration pollution will adversely affect crop growth, which can cause crop yield losses. Therefore, it is urgent to recognize the situation of ozone pollution in the BTH region and quantitatively evaluate the crop yield losses caused by ozone pollution to develop more effective pollution prevention and control policies. However, the monitoring of ozone concentration in China started relatively late compared with some developed countries, and currently, long-time series data covering the BTH region cannot be obtained, which makes it difficult to evaluate the impact of ozone on crop yield. Therefore, a new method (WRFC-XGB) was proposed in this study to establish a high-precision near-surface O₃ concentration dataset covering the whole BTH region from 2014 to 2019 by integrating the Weather Research and Forecasting with Chemistry (WRF-Chem) model with the extreme gradient boosting (XGBoost) machine learning algorithm. Through verification with ground observation station data, the results of WRFC-XGB are satisfactory, and R² can reach 0.78-0.91. Compared with other algorithms, the accuracy of the near-surface ozone concentration dataset is greatly improved, which can be used to estimate the impact of surface ozone on crop yield. Based on this dataset, the yield loss of winter wheat, rice, and maize caused by O₃ pollution was estimated by using the response equation of the relative yield and ozone dose index. The results showed that the total yield losses of winter wheat, rice and maize from 2014 to 2019 were 2659.21 million tons, 49.23 million tons and 1721.56 million tons due to ozone pollution in the BTH region, respectively, and the highest relative yield loss of crops caused by O₃ pollution could be 29.37% during 2014-2019, which indicated that the impact of ozone pollution on crop yield cannot be ignored, and effective measures need to be developed to control ozone pollution, prevent crop production loss, and ensure people's food security.

Keywords: Crop yield loss; Machine learning; Ozone; WRF-Chem.