Scholarlink

<p>Air pollution has become a global priority due to rapid development and its impact on public health and the environment. Predicting trace gas concentrations in the atmosphere helps legislators anticipate future conditions and take timely action. This study compares two modeling approaches: The Nonlinear Autoregressive Neural Network with Exogenous Inputs (NARX) and the Eulerian multiscale model. Measurement data were obtained from the Moroccan General Directorate of Meteorology, and both models were tested for their ability to model and predict ambient ozone concentrations. NARX delivered higher short-term accuracy, with a correlation coefficient of 94.21%, RMSE of 9.16&#xa0;µg/m³, and MAE of 7.33&#xa0;µg/m³. The Eulerian multiscale model better captured long-term trends and peak values but struggled with low concentrations and required more computational resources. Overall, NARX is well-suited for real-time, site-specific forecasting, while the Eulerian multiscale model is more appropriate for regional, long-term studies. Their complementary strengths suggest that hybrid modeling could enhance predictive accuracy and flexibility for air quality management.</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

Data-driven versus deterministic modeling for atmospheric trace gas prediction: insights from NARX and CHIMERE models

  • 期刊论文

  • 发表日期：

    2026年2月27日

查看全文
  • Anas Adnane,
  • Amine Ajdour,
  • Radouane Leghrib

摘要

Air pollution has become a global priority due to rapid development and its impact on public health and the environment. Predicting trace gas concentrations in the atmosphere helps legislators anticipate future conditions and take timely action. This study compares two modeling approaches: The Nonlinear Autoregressive Neural Network with Exogenous Inputs (NARX) and the Eulerian multiscale model. Measurement data were obtained from the Moroccan General Directorate of Meteorology, and both models were tested for their ability to model and predict ambient ozone concentrations. NARX delivered higher short-term accuracy, with a correlation coefficient of 94.21%, RMSE of 9.16 µg/m³, and MAE of 7.33 µg/m³. The Eulerian multiscale model better captured long-term trends and peak values but struggled with low concentrations and required more computational resources. Overall, NARX is well-suited for real-time, site-specific forecasting, while the Eulerian multiscale model is more appropriate for regional, long-term studies. Their complementary strengths suggest that hybrid modeling could enhance predictive accuracy and flexibility for air quality management.