Scholarlink

<p>A significant decreasing trend of Arctic stratospheric ozone has been observed since 2019, with the first reported ozone hole in the Arctic Stratospheric Polar Vortex (SPV) in 2020, raising concerns for humanity. This underlines that it is essential to develop an algorithm capable of predicting Arctic ozone levels, preferably using minimal computing resources. This study presents a novel approach for ozone prediction based on the morphological and dynamical properties of the SPV utilizing a explainable machine learning approach. XGBoost exhibits good agreement with the observations, achieving an <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_24379_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> score of 0.80 and a correlation of 0.91. The algorithm accurately predicts the daily and seasonal patterns of ozone variations. It successfully captures the pattern of the lowest recorded ozone levels in 2020, though it overestimates ozone values by approximately 20 Dobson units. Moreover, in some years the predicted ozone values also show a strong alignment with the observations. Notably, the algorithm relies solely on physics based features of the SPV to predict chemical ozone loss, demonstrating the potential of dynamical parameters in predicting the ozone variability. It could serve as a tool for projecting future Arctic ozone variability by utilizing input from climate models that lack interactive chemistry.</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

A novel approach to predict the arctic stratospheric ozone from stratospheric polar vortex dynamics using explainable machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月20日

查看全文
  • Anish Kumar,
  • Joyjit Mandal,
  • Sina Mehrdad,
  • Christoph Jacobi

摘要

A significant decreasing trend of Arctic stratospheric ozone has been observed since 2019, with the first reported ozone hole in the Arctic Stratospheric Polar Vortex (SPV) in 2020, raising concerns for humanity. This underlines that it is essential to develop an algorithm capable of predicting Arctic ozone levels, preferably using minimal computing resources. This study presents a novel approach for ozone prediction based on the morphological and dynamical properties of the SPV utilizing a explainable machine learning approach. XGBoost exhibits good agreement with the observations, achieving an \(R^2\) score of 0.80 and a correlation of 0.91. The algorithm accurately predicts the daily and seasonal patterns of ozone variations. It successfully captures the pattern of the lowest recorded ozone levels in 2020, though it overestimates ozone values by approximately 20 Dobson units. Moreover, in some years the predicted ozone values also show a strong alignment with the observations. Notably, the algorithm relies solely on physics based features of the SPV to predict chemical ozone loss, demonstrating the potential of dynamical parameters in predicting the ozone variability. It could serve as a tool for projecting future Arctic ozone variability by utilizing input from climate models that lack interactive chemistry.