Scholarlink

<p>This study investigates the temporal variations of ozone (O₃), carbon monoxide (CO), sulfur dioxide (SO₂), and particulate matter (PM₂.₅) concentrations in Brasília, Brazilian Midwest, focusing on the critical dry season and forest fire periods. Utilizing real data from 2000 to 2018, nonparametric trend tests (Mann-Kendall and Pettitt) and Principal Component Analysis (PCA) were applied to evaluate pollutant levels and their relationships with meteorological factors. Machine learning (ML) models were employed to predict O₃ concentrations. Random Forest and XGBoost models achieved good predictive results with MSE scores of 20.57 and 18.34, MAE 3.22 and 3.11, R² 0.69 and 0.72, and SMAPE 17.55 and 16.89, respectively. XGBoost model highlighted relative humidity as the most important feature, followed by CO, SO₂, wind speed, air temperature, PM₂.₅, and wind speed. While no significant long-term trends were detected, CO concentrations showed an upward tendency during the dry season, and O₃ and SO₂ were positively correlated with wind speed but negatively with humidity and precipitation. These findings underscore the transient influence of meteorological systems on pollutant levels and demonstrate the potential of ML models for advancing air quality research by complementing traditional monitoring systems and providing actionable insights for policymakers. The study provides actionable insights for developing targeted air quality management strategies and expanding monitoring networks to protect public health in Brasília.</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

Meteorological Influences and Machine Learning Approaches in Assessing Air Quality in Brasília

  • 期刊论文

  • 发表日期：

    2025年4月16日

查看全文
  • Amaury de Souza,
  • Marcel Carvalho Abreu,
  • Elias Silva de Medeiros,
  • Marcos Antônio Alves,
  • José Francisco de Oliveira Júnior,
  • Sneha Gautam

摘要

This study investigates the temporal variations of ozone (O₃), carbon monoxide (CO), sulfur dioxide (SO₂), and particulate matter (PM₂.₅) concentrations in Brasília, Brazilian Midwest, focusing on the critical dry season and forest fire periods. Utilizing real data from 2000 to 2018, nonparametric trend tests (Mann-Kendall and Pettitt) and Principal Component Analysis (PCA) were applied to evaluate pollutant levels and their relationships with meteorological factors. Machine learning (ML) models were employed to predict O₃ concentrations. Random Forest and XGBoost models achieved good predictive results with MSE scores of 20.57 and 18.34, MAE 3.22 and 3.11, R² 0.69 and 0.72, and SMAPE 17.55 and 16.89, respectively. XGBoost model highlighted relative humidity as the most important feature, followed by CO, SO₂, wind speed, air temperature, PM₂.₅, and wind speed. While no significant long-term trends were detected, CO concentrations showed an upward tendency during the dry season, and O₃ and SO₂ were positively correlated with wind speed but negatively with humidity and precipitation. These findings underscore the transient influence of meteorological systems on pollutant levels and demonstrate the potential of ML models for advancing air quality research by complementing traditional monitoring systems and providing actionable insights for policymakers. The study provides actionable insights for developing targeted air quality management strategies and expanding monitoring networks to protect public health in Brasília.