Scholarlink

<p>Dust storms are a major environmental hazard in arid regions, significantly affecting air quality, public health, and infrastructure. Accurate forecasting is essential, particularly in the Middle East, where dust activity is driven by regional meteorological and land surface conditions. While machine learning has been increasingly used for dust storm prediction, the combined use of surface meteorological observations and high-resolution satellite-derived aerosol products remains underexplored. The main objective of this study is to evaluate whether integrating MODIS MAIAC Aerosol Optical Depth (AOD, Optical_Depth_047) with surface meteorological data from the NOAA Integrated Surface Dataset improves the predictive accuracy and calibration of machine learning models for daily dust storm detection in Al-Ahsa, Saudi Arabia (2015–2024). To achieve this, two modelling frameworks were developed: Model 1, which used meteorological variables only (temperature, wind speed, and visibility), and Model 2, which incorporated AOD. Five algorithms Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and Multilayer Perceptron were assessed using temporal validation, k-fold cross-validation, and probabilistic calibration. Random Forest achieved the best performance, with high predictive accuracy and improved calibration when AOD was included (Brier score reduced from 0.087 to 0.080). These findings confirm the dominant role of meteorological variables while showing that AOD enhances probabilistic reliability, providing a robust framework for operational dust forecasting in arid regions.</p> Graphical Abstract <p></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

Machine Learning-Based Prediction of Dust Storms in Al-Ahsa, Saudi Arabia, Using NOAA Meteorological and MODIS Aerosol Data

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月6日

查看全文
  • Sarah Albugami

摘要

Dust storms are a major environmental hazard in arid regions, significantly affecting air quality, public health, and infrastructure. Accurate forecasting is essential, particularly in the Middle East, where dust activity is driven by regional meteorological and land surface conditions. While machine learning has been increasingly used for dust storm prediction, the combined use of surface meteorological observations and high-resolution satellite-derived aerosol products remains underexplored. The main objective of this study is to evaluate whether integrating MODIS MAIAC Aerosol Optical Depth (AOD, Optical_Depth_047) with surface meteorological data from the NOAA Integrated Surface Dataset improves the predictive accuracy and calibration of machine learning models for daily dust storm detection in Al-Ahsa, Saudi Arabia (2015–2024). To achieve this, two modelling frameworks were developed: Model 1, which used meteorological variables only (temperature, wind speed, and visibility), and Model 2, which incorporated AOD. Five algorithms Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and Multilayer Perceptron were assessed using temporal validation, k-fold cross-validation, and probabilistic calibration. Random Forest achieved the best performance, with high predictive accuracy and improved calibration when AOD was included (Brier score reduced from 0.087 to 0.080). These findings confirm the dominant role of meteorological variables while showing that AOD enhances probabilistic reliability, providing a robust framework for operational dust forecasting in arid regions.

Graphical Abstract