Spatiotemporal variations in meteorological influences on ambient ozone in China: A machine learning approach

文献类型： 外文期刊

第一作者： Li, Tao

作者： Li, Tao;Lu, Yichen;Zhan, Yu;Deng, Xunfei;Zhan, Yu

作者机构：

关键词： Ozone; Meteorological influence; Spatiotemporal variation; Random forest; Variable importance

期刊名称：ATMOSPHERIC POLLUTION RESEARCH （ 影响因子：4.5； 五年影响因子：4.6 ）

ISSN： 1309-1042

年卷期： 2023 年 14 卷 4 期

页码：

收录情况： SCI

摘要： Considering the increase in ambient ozone (O-3) levels with harmful health effects, this study aims to evaluate the spatiotemporal variations in meteorological influences on the daily maximum 8-h average O-3 concentrations ([O-3](MDA8)) across China. Leveraging the high capacity of the random forest in simulating complicated relationships between the predictor variables and [O-3](MDA8), we proposed a new method (named LVIG) to derive local variable importance from the global model (i.e., the random forest) for specific locations and months. On the basis of the LVIG results, the [O-3](MDA8) in the northern China was more associated with the evaporation and temperature, while the [O-3](MDA8) in the southern China was more associated with the relative humidity and sunshine duration. For the whole China, relative humidity was more influential during April to August, while evaporation, temperature and sunshine duration exhibited higher importance from November to February. The varying patterns of the meteorological influences could be explained by the Liebig law of the minimum, i.e., the limiting factors were the driving factors. Compared to the method of building multiple (geographically weighted) local models, the LVIG method gave more stable and specific estimates of local variable importance. As a generic method, the LVIG would be potentially applied in a wide range of fields.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序