Scholarlink

<p>Machine learning models have been extensively employed for predicting ground-level ozone (O<sub>3</sub>) in China, primarily focusing on short-term (1–3&#xa0;days) forecasts of O<sub>3</sub> pollution. This study established a dynamical–statistical model employing Random Forest (RF) and evaluated its capacity in forecasting O<sub>3</sub> pollution at a seasonal timescale. Observations indicated O<sub>3</sub> pollution was most frequent from April to September between 2014 and 2023 in Central-East China, with the monthly mean MDA8 O<sub>3</sub> exceeding 100&#xa0;μg&#xa0;m⁻<sup>3</sup> for the North China Plain (NCP), the Yangtze River Delta (YRD) and the Fen-Wei Plain (FWP). The seasonal forecast model predicted first-month monthly mean O<sub>3</sub> concentrations with the lowest NMB from the current month to the following three months. The RF model was tested using several statistical criteria and found to be reliable. In test datasets for NCP, YRD, and FWP, R<sup>2</sup> ranged from 0.68 to 0.77, with MAE and RMSE ranging from 16.3 to 17.2 and 21.2 to 22.1&#xa0;μg&#xa0;m<sup>−3</sup>. From April to September 2023, independent hindcasts showed that approximately 77% of cities had NMBs less than 10%. Our investigation showed that the RF model could forecast O<sub>3</sub> monthly mean concentrations one month in advance, greatly improve early warning and fortify systems for preparing for mitigation actions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Seasonal Prediction of Ozone Pollution in Central-East China Using Machine Learning

  • 期刊论文

  • 发表日期：

    2025年4月14日

查看全文
  • Sheng Yang,
  • Xin Hao,
  • Nan Li,
  • Hong Liao,
  • Jiandong Li

摘要

Machine learning models have been extensively employed for predicting ground-level ozone (O3) in China, primarily focusing on short-term (1–3 days) forecasts of O3 pollution. This study established a dynamical–statistical model employing Random Forest (RF) and evaluated its capacity in forecasting O3 pollution at a seasonal timescale. Observations indicated O3 pollution was most frequent from April to September between 2014 and 2023 in Central-East China, with the monthly mean MDA8 O3 exceeding 100 μg m⁻3 for the North China Plain (NCP), the Yangtze River Delta (YRD) and the Fen-Wei Plain (FWP). The seasonal forecast model predicted first-month monthly mean O3 concentrations with the lowest NMB from the current month to the following three months. The RF model was tested using several statistical criteria and found to be reliable. In test datasets for NCP, YRD, and FWP, R2 ranged from 0.68 to 0.77, with MAE and RMSE ranging from 16.3 to 17.2 and 21.2 to 22.1 μg m−3. From April to September 2023, independent hindcasts showed that approximately 77% of cities had NMBs less than 10%. Our investigation showed that the RF model could forecast O3 monthly mean concentrations one month in advance, greatly improve early warning and fortify systems for preparing for mitigation actions.